WAS vs JBoss, WebLogic, Tomcat (year 2015)

IBM WebSphere Application Server
competitive positioning
vs. JBoss EAP, Tomcat, TomEE, tc Server
Roman Kharkovski
IBM, Executive IT Specialist
kharkovski@us.ibm.com
December 11, 2015 © 2015 IBM Corporation

2
Leverages 100+
OSS Packages
Leverages 100+
OSS Packages
Leverages 30
OSS Packages
Leverages 40+
OSS Packages
MQ
3,000+ IBM developers involved in OSS projects. IBM leads 80+ and contributes to 350+ OSS projects.
http://www.ibm.com/developerworks/opensource

3
Source: Gartner, Market Share Analysis: Enterprise Software Market Share, Worldwide. Published March , 2015
2014
($B)
YTY
growth
rank share growth rank share growth rank share growth
BPM 2.5 6.4% # 1 28.5% 1.4% # 3 7.9% 2.3% # 22 0.2% -
ESB 2.77 10.4% # 1 29.0% 7.5% # 2 21.6% 2.6% # 16 0.4% 36.6%
MOM 1.35 8% # 1 75.0% 6.5% - - - # 9 0.3% 38.9%
MFT Suites 0.7 15.6% # 1 31.3% 11.4% # 13 0.9% - - - -
TP Monitors 1.86 2.8% # 1 85.2% 4.% # 2 10.0% -5.9% - - -
Appliances AIM 0.8 15.2 # 1 18.5% 7.1% # 3 0.8% 1.5% - - -
B2B 0.9 3.8% # 1 21.2% 0.7% - - - - - -
App Servers 5.4 14.6% # 2 28.5% 10.2% # 1 33.5% 1.6% # 5 2.6% 36.6%
Portals 1.8 1.4% # 2 24.4% -8.5% # 3 21.1% 2.4% # 10 0.8% 30.2%
Svc Governance 6.2 17.5% # 1 11.2% 6.1% # 2 10.6% 1.5% - - -
Other AIM 4.9 5.1% # 2 28.2% 30.0% # 17 0.6% -17% # 28 0.2% 32.5%
According to Gartner, IBM holds #1 position in the
middleware software for the past 13 years

4
Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors
with the highest ratings. Gartner research publications consist of the opinions of Gartner's research organization and should not be construed as statements of fact.
Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose
Magic Quadrant for On-Premises
Application Platforms
Daniel Sholler, Yefim V. Natis,
Massimo Pezzini, Kimihiko Iijima,
Jess Thompson, Ross Altman
June 27, 2013
This Magic Quadrant graphic was published by Gartner, Inc.
as part of a larger research note and should be evaluated in
the context of the entire URL
“New and composite on-
premises applications need a
complex array of runtime
technologies and development
capabilities.”
Source: Gartner (June 2013)
IBM named a leader in the Magic Quadrant for On-
Premises Application Platforms

5
Technical Overview
Features, functions, speeds and feeds…

6
What is “new” in WAS
v8.5(2012)  Improved performance, security, etc.
 Java EE 6 support, JDK 7
 Intelligent management features (used to be WVE) – health management,
application versioning, dynamic clusters, traffic throttling, SLA control, etc.
 Added batch (used to be separate product)
 New Liberty Profile (developer focus)
 Improved OSGi support and SCA 1.1
 Pluggable JDK (choice of v6 or v7)
 Configuration checkpoints (full, delta, compare, restore, etc.)
 WAS Express limit per server raised up to 480 PVU
 HTTP clustering for WAS Base for up to 5 JVMs
 Socket based license option for WAS Base (since v7)
v8.5.5(2013)
 Improved performance, security, JMS, etc.
 Troubleshooting (XCT, memory leak policies, HPEL, ISA Data Collector)
 WXS bundled with all WAS editions
 Liberty Profile with enhanced programming model (Web profile + JMS, MDB,
JAXWS, clustering, etc. – production focus)
 ODR merged into IHS and Apache server plugin
 Service mapping (from WESB)
 Free Liberty Core license for ISV applications
 HTTP clustering for WAS Express, Base and Liberty Core is now unlimited
 Fixed Term License option for all editions of WAS (like annual subscription)

7
What is new in WAS Liberty Profile
2H’2014
 Improved performance, security, etc.
 Auto Scaling and Dynamic Routing
 Partial Java EE 7 (Servlet 3.1, WebSocket 1.0, Concurrency 1.0, JSON-P 1.0)
 Improved v2v and competitive Migration Toolkit
 Web-based SSO for applications with OpenID 2.0
 Support for CouchDB
 REST connector for non-Java clients
 Support for Enterprise Web Services (JSR 109 MR)
 A number of beta features (SIP, JMS 2.0, JAX-RS 2.0, JDBC 4.1, JPA 2.1,
Batch, WebRTC, bean validation 1.1, EJB 3.2 lite, etc.)
 and more…
1H’2015
 Production ready support and certification for Java EE 7 and Java SE 8
 Improved Admin Center (tagging, searching, monitoring, scalability, config)
 Improved developer tools (remote debugger, repository integration)
 Improved support for IBM Bluemix PaaS
 Docker images and support
 SPNEGO, SAML (beta)
 Log collector and analytics (beta)
 No-charge Liberty Base for production (up to 2GB Java heap per organization)
 “2 for 1” licenses for 6 months on SoftLayer
 Monthly pricing option (for most of IBM software)
First production server to do so!

8
WAS Liberty Profile
+
WAS Developer Tools
for Eclipse
Unzip install / deploy
IM or unzip to install. Option to
deploy “server package” of app +
config + required subset of server
runtime for highest density deploy
User Extensions
Add custom features and
integrate 3rd party
components via Liberty
extensions interface
Fidelity to full profile
Same reliable containers & QOS.
Develop on Liberty profile and
deploy to Liberty or full-profile WAS
Integrated tools
Powerful tools in free WDT
Eclipse feature, maven / ant /
Jenkins integration, migration
tools, Bluemix integration
Lightweight cluster mgmt
Liberty servers can join a lightweight
cluster for workload balancing and high
availability
Dynamic Server Profile
No server restarts required. Configured by app at
a fine-grained level. Install new features from
repository (local or remote) with no restart
Start fast, run efficiently
Starts in <3s; Mem footprint <70MB;
(TradeLite benchmark)
Cloud enabled
Hosted on IBM Bluemix PaaS.
Buildpacks for other cloud providers
“Developer First” Focus
Simplified to one file, shareable XML server
configuration with Zero V2V migration.
Continuous delivery of new function – new
beta every month. Free for development
Small Download
~60MB for Web Profile features,
~100MB for full profile
Java EE 7 Certified
and Java EE Web Profile 6 & 7
certified
What is WAS Liberty Profile
Free for development and up to 2 GB for Liberty Base production

9
WAS Liberty zOS
WAS ND Liberty
webProfile-6.0
WAS Liberty Core
WAS Liberty
Liberty v8.5.5.6 was the first Java EE 7 production server
collectiveController-1.0clusterMember-1.0 scalingController-1.0 scalingMember-1.0 dynamicRouting-1.0
webSocket-1.1
webSocket-1.0servlet-3.1
jsp-2.3
jsf-2.2
ejbLite-3.2 jdbc-4.1
jndi-1.0
appSecurity-2.0
managedBeans-1.0
ssl-1.0
beanValidation-1.1
cdi-1.2
jpa-2.1
jaxrs-2.0
jaxrsClient-2.0
el-3.0
jsonp-1.0
mongodb-2.0
wsSecurity-1.1
wmqJmsClient-2.0
wasJmsServer-1.0
jmsMdb-3.2
wasJmsClient-2.0jaxws-2.2
jaxb-2.2
wasJmsSecurity-1.0
jca-1.7
couchdb-1.0
jcaInboundSecur-1.0
mdb-3.2
jms-2.0 ejb-3.2
j2eeManagement-1.1
ejbPersistentTimer-3.2
ejbRemote-3.2
jaspic-1.0
jacc-1.0
batch-1.0
appClientSupport-1.0 javaeeClient-7.0
ejbHome-3.2
openid-2.0
openidConnectServer-1.0
openidConnectClient-1.0
spnego-1.0
osgiAppIntegration-1.0
wab-1.0
concurrent-1.0
collectiveMember-1.0
restConnector-1.0
sessionDatabase-1.0
ldapRegistry-3.0
webCache-1.0
distributedMap-1.0
osgiConsole-1.0
json-1.0
timedOperations-1.0monitor-1.0
oauth-2.0
blueprint-1.0
adminCenter-1.0
serverStatus-1.0
eventLogging-1.0
requestTiming-1.0
javaMail-1.5
zosSecurity-1.0 zosTransaction-1.0 zosLocalAdapters-1.0zosWlm-1.0zosConnect-1.0
batchManagement-1.0
Newin8.5.5.6

10
Liberty WAS WebLogic TomEE Tomcat tc Server JBoss
Java EE 7 Beta WildFly
Java EE Web profile 7 Beta WildFly
Java EE Web profile 6
Java SE 7 and 8 8 in Beta
Servlet, JSP, JSF MyFaces JSF
JDBC
Java Persistence API (JPA)
Java Message Service (JMS) * TomEE+ RabbitMQ 1.1
Java Transaction API (JTA)
Bean validation
Java Management Extensions (JMX)
Java API for XML Web Services (JAX-WS) * TomEE+
Context Dependency Injection (CDI)
Java API for RESTful Web Services (JAX-RS) TomEE+
OSGi
EJB lite
EJB full
WebRTC
WebSocket (JSR 356) Beta
JSONP
Oauth
Concurrency Utilities for Java EE (JSR 236) WildFly
Batch API (JSR 352) Beta Spring Batch WildFly
JNDI Read-only Read-only
SAML
WS-Notification
WS-Policy
WS-Trust
WS-ReliableMessaging
Session Initiation Protocol (SIP) *
Portlet API dev only
WS-Addressing
RMI-IIOP *
Java Connector Architecture (JCA) * TomEE+
Java Auth. & Authoriz. Service (JAAS)
JACC and JASPIC
Excellent
Good
Limited
No support
API support

11
Liberty
8.5.5.7
WAS
8.5.5.6
WASND
8.5.5.6
Tomcat
8.0.26
tcServer
3.1.2
JBoss
EAP 6.4
WLS
12.2.1
Java EE 7 Beta Beta SoD
Java EE 6 Web Profile TomEE
JDK 1.7 and 1.8 1.7 1.7 3rd party 3rd party RHEL only
Performance
Security
Transaction management TomEE
Messaging Engine TomEE+
Caching engine included (IMDG) WXS WXS WXS Infinispan
Admin GUI
Single
server
Troubleshooting, profiling, tuning tools
Admin scripting
Simple V2V migration and upgrades
Log analytics Beta
Dynamic clustering and auto-scaling ND
SLA enforcement and monitoring for requests
Application versioning
Automated server health management
EJB and JMS clustering and failover
HTTP plugin with WLM and HTTPSession failover
Dynamic configuration updates (avoid restarts)
Simple install, lightweight runtime, small footprint
Cloud (Public / Private / Hybrid) 3rd party
Appliance IPAS IPAS IPAS
Free sw included (WLM, HTTPD, LDAP, DBMS) ND
Platform certifications (OS, HW, DBMS, Adapters)
WW support (local language, local hours)
Excellent
Good
Limited
Very Limited
No support

12
WAS
classic
WAS
Liberty
Oracle
WLS
JBoss
EAP
Apache
Tomcat
Basic web application security Yes Yes Yes Yes Yes
Java EE Security Standards Yes Yes Yes Yes no
Role administrative security (who can do what) Yes no Yes no no
Resource administrative security (who can manage what) Yes no Yes Yes no
Audit and track changes made to the server configuration Yes no Yes Some no
LDAP support and compatibility Yes Yes Yes Some Some
Federated User Registry Yes Yes Yes no no
SPNEGO Web Inbound Yes Yes Yes Yes no
Kerberos Yes no Yes Yes no
OAUTH Yes Yes Yes Yes no
SAML Yes Yes Yes Yes no
OpenID Yes Yes no no no
OpenID Connect Yes Yes no no no
Keys and Certificate Management Yes no Yes no no
Multiple Security Configurations Yes no Yes no no
DB2 Trusted Context for Identity propagation Yes Yes no no no
Secure Engineering Accreditation OTT-PS Yes Yes no no no
Encryption Standard FIPS 140-2, 800-131a Yes Yes no Some Some
Security

13
So you say: “I simply use Tomcat”?
• Most people use a *lot* more than just Tomcat. What about:
• Once you add all of these (and more), how do you make it work together? (i.e.
3rd party CDI wont work with 4th party JTA, etc.)
• What is the performance of all of the above with all the JAR scanning going on?
• How do you test/manage/secure/debug all of this?
• Are you coding to standards? Are you coding your app or building your own
Java EE server?
• Lack of API support in Tomcat leads to lost development
productivity and additional expense for integration and testing
REST
SOAP
JPA
WS-*
JTA
Security
JMS
CDI
Bean
validation
OSGI
JTA
EJB
WebRTC
Batch
Concurre
ncy
Oauth
SAML
Other
Java EE?
vs.

14
Network
Deployment
(classic + Liberty)
JavaWeb
(Servlet/JSP,JDBC)
JavaEE(WS*,JMS,EJB,JPA,
CDI,JAXB,JTA,JCA,Batch,etc.)
OSGi,SIP,SCA,SAML,
Kerberos,COBOL/z,etc.
Hundreds of serversHandful of servers
Programmingmodel(APIs)
Base
(classic + Liberty)
Express
(classic + Liberty)
WebLogic
Standard Enterprise Suite
WebSphere
Liberty Core
WebSphere
Scale
Server capability map
JBoss
EAP
Tomcat
TomEE
tc Server

15
Java servers from Developer point of view
Liberty
8.5.5.7
WAS
8.5.5.6
Tomcat
8.0.26
TomEE+
1.7.24
Jetty
9.3.2
Glass
Fish 4.1
Web
Logic
12.2.13
WildFly
9.0.1
JBoss
EAP 6.4
Server stop+start5 4.9 sec 34.1 sec 5.5 sec 11.2 sec 3.1 sec 9.4 sec 22.3 sec 10.2 sec 9.2 sec
App redeploy5 1.2 sec 6.1 sec 2.3 sec 2.5 sec 2.2 sec 2.5 sec 5.0 sec 1.2 sec 1.2 sec
RAM5 59 MB 175 MB 125 MB 236 MB 102 MB 376 MB 451 MB 269 MB 430 MB
Download size1 11 to 94 MB 3 GB 10 MB 48 MB 10 MB 103 MB 211 MB 127 MB 158 MB
Size installed1 15-123 MB 2.6 GB 17 MB 52 MB 12 MB 214 MB 595 MB 159 MB 174 MB
Size per instance 0.5 MB 40 MB 0.4 MB 0.4 MB 0.4 MB 96 MB 6 MB 1.5 MB 1.2 MB
Dev. Install6 5 sec 30 min 2 sec 3 sec 1 sec 5 sec 10 min 5 sec 5 sec
# of config files 1+ 100+ 8+ 12+ 20+ 14+ 22+ 16+ 16+
Dynamic config2 99% 80% 20% 20% 20% 20% 80% 60% 60%
IDE
Configuration Editor Eclipse UI Browser UI None None None Browser UI Browser UI Browser UI Browser UI
DevOps
Java EE Java EE 7 Java EE 6+
JSP/
Servlet
Java EE 6
Web Prof.
JSP/
Servlet
Java EE 7 Java EE 7 Java EE 7 Java EE 6
Free Dev. License IBM IBM Apache 2.0 Apache 2.0 EPL 1.0 CDDL 1.1 Oracle LGPL 2.1 LGPL 2.1
Free Dev. Support IBM7 IBM7 Self Self Self Self $ Self Red Hat8
Excellent
Good
Limited
Very Limited
No support
Maven, Jenkins, Ant, Chef and other DevOps tools are supported with some minor differences
Eclipse, IntelliJ IDEA, NetBeans are supported with some minor differences

16
OSGi support
• First surfaced to applications in a WAS v7 Feature Pack
• Modular development, deployment and management
• Blueprint (Standardized Spring Component Model)
• Web applications (Java EE 5)
• Remote Services and Heterogeneous Assembly (SCA)
• Included in the WAS Base in v8 and continually extended:
• Java EE 6 Web technologies
• Post-deployment configuration
• Performance metrics
• In-place Update
• Application Extension
• Modular EJB
• Blueprint Role-based Security
• OSGi Applications Web Console
• Liberty Profile support
• WAS classic and WAS Liberty have good OSGi support
Tomcat and JBoss do not support OSGi

17
WebSphere release-to-release performance increases due to
software and hardware improvements
EjOPS/core
As per SPEC Published Data as of 2/18/2015: http://www.spec.org/jEnterprise2010/results/jEnterprise2010.html
SPECjEnterprise2010 benchmark results
126.7
149.4
226.7
292.6 307.9
524.6
606
688
754
823
939
WAS
7.0.0.5 (8
core x86)
WAS
7.0.0.9 (8
core x86)
WAS
7.0.0.9 (8
core x86)
WAS 8.0
(8 core
x86)
WAS 8.0
(12 core
x86)
WAS 8.5
(12 core
x86)
WAS 8.5
(16 core
x86)
WAS
8.5.5.4 (28
core x86)
WAS 8.5
(16 core
Power 7)
WAS 8.5.5
(16 core
Power 7+)
WAS
8.5.5.2 (24
core Power
8 s824)
January
2010
April2014
SPEC and SPECjEnterprise 2010 are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 02/18/2015 IBM SPECjEnterprise results mentioned are 1013.40 EjOPS, 1194.80 EjOPS, 1813.37 EjOPS,
2341.12 EjOPS, 3694.35 EJOPS, 6295.46 EjOPS, 9696.43 EjOPS, 19282.14 EjoPS, 12,066.73 EjOPS, 13,161.07 EjOPS and 22,543.34 EjOPS published on Jan 2 2010, Feb 25 2010, Apr 27 2010, Jun 20 2011, Jun 17 2011, Apr 26 2012, Nov 14 2012,
Feb 18, 2015, Mar 6 2013, Apr 22 2013 and Apr 22, 2014 respectively

18
SPECjEnterprise2010
Comparison of IBM vs. Oracle
performance JOPS per core starting
from 2011
W
LS
12c
on
T5-2
(Jan'14)
W
LS
12c
on
T5-8
(Sep'13)
W
LS
11g
on
T5-8
(M
ar'13)
W
LS
11g
on
Sun
x86
(Feb'12)
W
LS
11g
on
Sun
x86
(Jul'11)
W
LS
11g
on
T4-4
(Aug'11)
W
LS
11g
on
Dellx86
(Apr'11)
JOPS/core 532.30 457.14 448.61 519.39 452.29 313.32 298.67
WAS 8.5.5.2 on Power8 (Apr'14) 939.31 1.76 2.05 2.09 1.81 2.08 3.00 3.15
WAS 8.5.5 on Power7+ (Apr'13) 822.57 1.55 1.80 1.83 1.58 1.82 2.63 2.75
WAS 8.5 on x3650 x86 (Nov'12) 606.03 1.14 1.33 1.35 1.17 1.34 1.93 2.03
WAS 8.5 on Power7+ (Sep'12) 681.39 1.28 1.49 1.52 1.31 1.51 2.17 2.28
WAS 8.5 on HS22 blade x86 (Apr'12) 524.62 0.99 1.15 1.17 1.01 1.16 1.67 1.76
WAS 8.5 on HS22 blade x86 (Jul'11) 307.86 0.58 0.67 0.69 0.59 0.68 0.98 1.03
WAS 8.5 on HS22 blade x86 (Jun'11) 292.64 0.55 0.64 0.65 0.56 0.65 0.93 0.98
1 even result
>1 IBM advantage
<1 Oracle advantage
More recent results
Morerecent
Benchmark results
SPECjEnterprise2010
Comparison of IBM WAS ND vs. Oracle
WLS Enterprise: $ cost per JOPS
starting from 2011 W
LS
12c
on
T5-8
(Sep'13)
W
LS
11g
on
T5-8
(M
ar'13)
W
LS
11g
on
Sun
x86
(Feb'12)
W
LS
11g
on
Sun
x86
(Jul'11)
W
LS
11g
on
T4-4
(Aug'11)
W
LS
11g
on
Dellx86
(Apr'11)
$/JOPS $131 $153 $251 $200 $175 $245
WAS 8.5.5 on Power7+ (Apr'13) $81 1.62 1.90 3.11 2.47 2.16 3.03
WAS 8.5 on x3650 x86 (Nov'12) $111 1.18 1.38 2.26 1.80 1.57 2.21
WAS 8.5 on Power7+ (Sep'12) $223 0.59 0.69 1.13 0.90 0.78 1.10
WAS 8.5 on HS22 blade x86 (Apr'12) $244 0.54 0.63 1.03 0.82 0.72 1.00
WAS 8.5 on HS22 blade x86 (Jul'11) $168 0.78 0.91 1.50 1.19 1.04 1.46
WAS 8.5 on HS22 blade x86 (Jun'11) $108 1.21 1.42 2.33 1.85 1.62 2.27
Morerecent
More recent results

19
IBM WebSphere 12 years of performance leadership
SPECjEnterprise2010
(1) SPEC and SPECjEnterprise2010 are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 04/04/2013 Oracle SUN SPARC T5-8 449 EjOPS/core SPECjEnterprise2010 (Oracle's WLS
best SPECjEnterprise2010 EjOPS/core result on SPARC). IBM Power730 823 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result), (2) Results from www.spec.org as of 04/29/2012 Oracle SUN SPARC T4-4 313 EjOPS/core
SPECjEnterprise2010 (Oracle's WLS best SPECjEnterprise2010 EjOPS/core result on SPARC). IBM Power780 681 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result), (3) Results from www.spec.org as of 11/14/2012 Oracle
SUN Fire X4170M3 519.39 EjOPS/core SPECjEnterprise2010 (Oracle's WLS best SPECjEnterprise2010 EjOPS/core result on Sandy Bridge). IBM WAS 8.5 System x3650 M4 Intel Sandy Bridge EjOPS/core (World Record SPECjEnterprise2010
EJOPS/core result) (4) Results from www.spec.org as of 04/29/2012 Oracle SUN Blade Server X6270 M2 452.285 EjOPS/core SPECjEnterprise2010). IBM Websphere HS 22 Blade 524.621 EjOPS/core.
EjOPS per processor core (i.e. transactions per core)
524
452
12 cores of Intel Westmere Xeon X5690 processor4
681
313
Oracle Sun SPARC T4-4 vs. IBM Power7 hardware2
606
519
16 cores of Intel Sandy Bridge Xeon E5-2690 processor3
 IBM held the most records in ECPerf
and was FIRST to publish
SPECj2001, SPECj2002, SPECj2004,
SPECjEnterprise2010
 WAS is 32% faster per core on latest
Intel Haswell at half the cost
compared to WebLogic1
 On latest Intel Haswell processors
WAS has the fastest per socker, per
core and biggest total EjOPS result
compared to WebLogic2
 WAS is 105% faster per core at almost
half the cost on Power7+ compared
to WebLogic on SPARC T53
939
457
Oracle Sun SPARC T5-8 vs. IBM Power7+ hardware1
Intel x64 Haswell (February 2015) 5
IBM: 688
Oracle: 522

20
WAS performance compared to JBoss
(1) SPEC and SPECjEnterprise2010 are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 07/30/2015: IBM Power730 939 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result),
(2) Since JBoss EAP 6.4 does not support Java EE 7, results for WildFly are provided
Transactions per core
JBoss EAP: 2,721
IBM Liberty: 2,831
TradeLight on x86 (Java EE 7, native OS) 2
803
2,027
SPECjEnterprise2010 (Java EE 6, native OS, WAS Full Profile) 1
1,214
1,574
DayTrader7 on x86_64 (Java EE 7, Docker) 2
 IBM held the most records and was first
to publish SPECj2001, SPECj2002,
SPECj2004, SPECjEnterprise2010 while
Red Hat has never published a single
result for JBoss EAP
 IBM holds World Record for # of
transactions per second per core with
SPECjEnterprise2010 workload
 WAS is up to 2.5 times faster than JBoss
EAP (just consider license, hardware,
power and cooling savings!)
 Many independent customer
benchmarks confirm WAS performance
advantage
 IBM is heavily investing in performance
optimizations of WAS for Docker
Containers and CloudFoundry
1,316
2,285
DayTrader7 on x86_64 (Java EE 7, native OS) 2
DayTrader3 on x86_64 (Java EE 6, native OS)
939
JBoss never published SPECjEnterprise2010
1.7x
2.5x
1.3x
1.04 times faster
∞x

21
WAS performance compared to Tomcat
(1) SPEC and SPECjEnterprise2010 are registered trademarks of the Standard Performance Evaluation Corporation. Results from www.spec.org as of 07/30/2015: IBM Power730 939 EjOPS/core (World Record SPECjEnterprise2010 EJOPS/core result),
Transactions per core
Tomcat: 4,491
IBM: 5,661
TradeLight on x86 (native OS) 2
3,824
8,232
SPECjEnterprise2010 (native OS, WAS Full Profile) 1
4,775
9,297
TradeLight on x86 (two Docker containers) 2
 IBM held the most records and was first
to publish SPECj2001, SPECj2002,
SPECj2004, SPECjEnterprise2010 while
Red Hat has never published a single
result for JBoss EAP
 IBM holds World Record for # of
transactions per second per core with
SPECjEnterprise2010 workload
 Liberty is up to 2.4 times faster than
Tomcat (just consider license, hardware,
power and cooling savings!)
 Many independent customer
benchmarks confirm WAS performance
advantage
 IBM is heavily investing in performance
optimizations of WAS for Docker
Containers and CloudFoundry
3,188
7,687
TradeLight on x86 (one Docker container) 2
TradeLight on x86 (four Docker containers) 2
939
Tomcat can not run SPECjEnterprise2010
2.4x
2.1x
1.9x
1.3 times faster
∞

22
Global transactions
• Enterprises often have to update data in:
• Databases (DB2, Oracle DB, MS SQL, Informix)
• TP monitors (CICS, Tuxedo)
• Messaging servers (IBM MQ, Tibco EMS, ActiveMQ)
• Caching servers (WXS), etc.
• What happens if there is a failure during the transaction commit?
• Tomcat and tc Server do not include a transaction manager
• Only support local transactions (with a single resource)
• Without a distributed transaction manager, there is no support for two-phase
commit for global transactions, where the scope is across multiple resources
• Writing business critical code without global transaction management is possible,
but is very complex (i.e. error prone and expensive)
• WAS and Liberty support distributed transactions with two-phase commit and
automatic data recovery in the event of a network failure, ensuring transaction
integrity

23
Cloud
The saga of “a lost million”
Transaction: Transfer $1M
1. Check funds availability in Account1
2. Withdraw funds from the Account1
3. Debit funds to the Account2
4. Update related systems with the right information
5. Write into the audit log for security and compliance reasons
Account 1
balance = $1M
Account 2
balance = $0
Account 1
balance = $1M
Account 2
balance = $1M
Account 1
balance = $0M
Account 2
balance = $0M
Account 1
balance = $0M
Account 2
balance = $1M
Initial state:
Bad outcome:
Bad outcome:
Good outcome:
Credit Card System
Account 1
balance = $1M
Account 1
balance = $1M
Internet Banking System
Account 1
balance = $1M
Account 1
balance = $1M
Audit Log
Account 1
balance = $1M
Failure
Failure
Failure
Failure
Failure
Failure
Failure
Failure
Failure
Failure
Failure Failure
Failure
Types of failures:
1. Power outage
2. Network error or outage
3. Software failure (OS, DBMS, etc.)
4. Hardware failure
5. Human error
6. Application error
7. DoS or other attack
8. Combination of any of the above

24
Automatic transaction log failover & recovery
WAS ND Full Profile provides failover of 2PC transactions
• WAS ND can be configured to store transaction logs for each server on shared
file system or in the HA RDBMS, which allows all peers to see all logs
• When a WAS ND cluster member fails, a peer is elected to recover the
Transaction Log from the failed server
• In Doubt Transactions from a failed Server are recovered very quickly
• Significantly faster and cheaper than hardware clustering (seconds vs. minutes)
JBoss, Tomcat and Liberty Profile do not provide similar capabilities,
which leads to longer recovery times and more admin labor

25
JBoss high availability issues
• Repeated application redeployments eventually require server restarts due to the
memory leak issue (Sun and Open JDK PermGen heap gets full and crashes
server)
• HTTP servers and JBoss App Server JVMs often need restart after configuration
updates
• Examples: changing the data source pool size or other settings, changing JMS
configuration, re-deploying an EAR multiple times (only WAR hot-deploy works, the EAR
hot deploy only works for the first couple of times, then causes out of memory errors and
JVM crash), etc.
• WAS does not need to be restarted as above updates are dynamic
• JON server uses database for configuration and monitoring data (Postgres or
Oracle)
• The DB must be made HA to avoid SPOF and this requires extra license and hw costs
• Ripple restart of application servers in a cluster is not provided. Administrator must
manually restart servers in a cluster one by one
• Transaction log, JMS queues and topics require manual failover effort in case of
server failure
• WAS failover is automated and takes seconds
• Application deployment causes service interruption
• WAS ND solves this by introducing the application versioning and graceful client transfer
High availability with JBoss may be difficult to achieve without
introducing significant redundancy and admin effort

26
Liberty Admin Center
A browser-based UI for deploying,
monitoring and managing Liberty single
servers and collectives
• Deploy
• Server package (runtime + server + apps)
• Monitor
• Performance and Health metrics
• Log Analytics [beta]
• Dashboard, Alerts and Notifications
[under investigation]
• Manage
• Browse, search, filter
• Tags and metadata
• Start / Stop / Restart
• Auto scaling (demo)
• Server Configuration [beta]
• Health Management [under investigation]
Tomcat does not provide similar capabilities,
which leads to more admin labor

27
27
IBM Confidential
Resource details, multiple target actions, asynchronous
Liberty server deployment

28
Beta
Liberty Monitoring and Log Analytics
Assess the health of your application server using
out of the box feature in Liberty to:
• Monitor your Application Infrastructure
• Identify the root cause of a problem
• Ensures High Availability of your application
• Plus wide set of additional Java monitoring tools
available in the IBM Java Health Center
Access the customizable graphs from
the Admin Center to:
• Monitor:
• Used Heap Memory
• Loaded Classes
• Active JVM Threads
• CPU Usage
• Liberty MXBeans
• Analyze data from:
• Access Log, Log Messages, FFDC &
Trace Messages

29
Principles of Liberty Collective
• Standards-based admin API
• Built on JMX (MBeans)
• Works with common tools (Jconsole, Jython, etc)
• Loosely-coupled
• Exploits Liberty composable server model
• App servers can easily and quickly be moved in and out of collective
• Tested up to 10,000 servers in a collective
• Distributed Cache Model
• Admin server (controller) is config/state cache
• Doubles as bi-directional JMX proxy
• Distributed Configuration
• App Server owns its own config
• App Servers cache sparse config and state in controller
• Scalable, Resilient Admin Domain
• Highly Available admin server (replica model)
• Agentless
• Docker images can be part of the collective and auto-scale (beta)
Tomcat does not provide similar capabilities => more admin labor

30
Liberty Collective: lightweight management at scale

31
BlueMix vs. OpenShift Online app auto-scaling for Java EE
IBM
BlueMix
Red Hat
OpenShift
Scaling metrics:
JVM Heap Yes no
Memory Yes no
Throughput Yes no
Response time Yes no
# of connections no Yes
User options:
Breach duration Yes no
Statistic window Yes no
Cooldown period for scaling in&out Yes no
Scale in & out instance count Yes no
Max & min instance count Yes Yes
=> Less admin work
and more
responsive
applications with
IBM BlueMix

32
WAS classic console command assistance
Automatic capture of administrative actions and generation of scripts to be replayed later
• While administrator
performs actions in the
admin GUI (start, stop,
deploy, create, etc.) all his
actions are automatically
written as Jython
command script for WAS
• This script can be
customized and executed
multiple times thus saving
time to create complex
administrative actions and
reducing the learning
curve
Liberty, Tomcat and JBoss do not offer comparable capabilities

33
How do customers really use JBoss EAP in production?
• Vast majority of JBoss EAP and WildFly customers are not using clustering
• Must tolerate lower quality of services ($$$)
and
• Red Hat customers are forced to purchase 3rd party management tools,
monitoring tools, configuration management tools, performance profilers, etc.
• 3rd party tools require license and support payments ($$$)
• 3rd party tools are not always in synch with the desired version of JBoss ($$$)
• 3rd party vendor viability poses risks ($$$)
and
• Most JBoss customers invest significant staff time to build home grown
scripting frameworks for JBoss management (a combination of shell scripting
and generation of JBoss XML files using XSLT, Java or other template
mechanism)
• Cost to develop, debug, maintain such scripts can be significant ($$$)
• New major versions of JBoss (major or minor) are not backwards compatible,
causing significant rework of home grown scripts and tools ($$$)
• WAS provides needed administrative tools out of the box at no extra cost
”One minute of system downtime can cost an organization anywhere from $2,500 to $10,000 per minute.
Using that metric, even 99.9% availability can cost a company $5 million a year” - The Standish Group

34
IBM Garbage Collection and Memory Visualizer (GCMV)
GCMV provides analysis and views of your applications verbose gc output. GCMV uses
a powerful statistical analysis engine which provides tuning recommendations in these
areas:
• Memory Leak Detection
• Detect Java heap exhaustion and memory leaks
• Detect "native" heap exhaustion and memory leaks
• Optimizing garbage collection performance
• Analyze output from different gc modes (optthruput, optavgpause, gencon, balanced )
• Compare output from multiple logs – side by side
• Determine gc overhead, detect long or frequent gc cycles and causes
• Recommend settings to avoid long or frequent gc cycles
• Recommend optimum gc policy
• Fine tuning of Java heap size
• Determine peak and average memory usage
• Recommend Java heap settings
• Flexible user interface makes it possible to carry out further analysis of the data and
to "drill down" into the causes of trends and export of data into .csv or jpeg
Oracle Java Mission Control (JMC) is free for development use only.
JMC does provide data visualization, but it does not make tuning
recommendations, nor does it compare various run results side by side. This
is a major usability issue.

36
IBM WebSphere Performance Tuning Toolkit (PTT)
PTT is designed to help users tune the performance of WAS using statistical
technology. The toolkit collects performance data and consolidates it into a
multidimensional data cube.
• Find potential performance problems
• PTT shows detailed status of system with easily understood charts and forms. Users can
analyze the performance data from various perspectives.
• PTT helps to find an error as soon as it occurs - monitor the servlet errors, transaction
rollback, transaction timeout, JDBC connection timeout, thread hung, etc.
• Accelerate performance tuning process
• User can tune many servers in one step in a centralized view by running tuning scripts
within the workbench, download or upload performance related settings manually or via
script
• Health Check
• PTT can detect the performance decline and take actions automatically based on
predefined rules. Rule engine detects the abnormal symptoms according to user defined
rules (with ability to create and edit existing rules)
• Operations to facilitate problems determination
• PTT can generate thread dump and heap dump for the JVM, enable trace settings,
extract the connection pool contents
• Report engine
• Online and offline analysis and reporting (generate, export and print report)
Those using WebLogic, JBoss and Tomcat must spend considerably more
effort finding all the right tuning variables. In these products the monitoring
data is scattered across multiple locations in the Admin GUIs or worse –
only available for custom JMX programs

38
IBM Monitoring and Diagnostic Tools for Java - Health Center
Health Center is a diagnostic tool for monitoring the status of a running JVM. It uses a
small amount of processor time and memory, and can open some log and trace files for
analysis:
• Monitoring a running Java application or recorded activity for offline analysis
• Very low performance overhead allows to connect to and monitor a live Java application
(or replay recorded activity), such as CPU, environment, IO, gc, locking, threads,
memory, method tracing with timings, etc.
• Save data from a monitored Java application, then reload the saved data later on, without
making a live connection. You can load data from multiple files by loading one file, then
appending more files.
• Viewing the data collected
• Displays the data collected using different views (graphical and tabular)
• Triggering dumps
• Trigger the JVM to generate System Dumps, Heap Dumps, and Java Dumps
• Troubleshooting
• The first step in troubleshooting is to view the log files that are produced by the Health
Center client and agent. Then read the information provided for some of the common
problems that you might encounter.
• Performance hints
• The Health Center agent has little effect on performance. You can improve the
performance of the Health Center client by reducing the amount of data collected or
displayed.
• You can use the Health Center API to write your own code for manipulating Health
Center data

41
Find performance problems fast
• Which request is slow? Which is hung? Why?
• Slow servlet requests are detected and a full diagnostic of the request is dumped to the log
• Hung servlet requests are detected, triggering creation of a set of javacores
• Use in your production environments to catch issues the first time they occur
TRAS0112W: Request websphere.servlet.service|DayTrader Web | TradeScenarioServlet(AAC9KLwFFXT_AAAAAAAAAAN) has been
running on thread 0000006b for 1549.460ms. The following stack trace shows what this thread is currently doing.
<stack trace>
The following table shows the events that have run during this request.
Duration Operation
1552.012ms + websphere.servlet.service | DayTrader Web | TradeScenarioServlet
0.014ms websphere.session.getAttribute | R-ObCtcDfR8Zd9riQEMCh6R | uidBean
30.714ms websphere.servlet.service | DayTrader Web | TradeAppServlet
30.456ms websphere.servlet.service | DayTrader Web | /quote.jsp
28.903ms websphere.servlet.service | DayTrader Web | /displayQuote.jsp
0.194ms websphere.datasource.psExecuteQuery | jdbc/TradeDataSource | SELECT t0.CHANGE1, t0.COMPANYNAME..
1520.695ms + websphere.servlet.service | DayTrader Web | TradeAppServlet
0.190ms websphere.datasource.psExecuteQuery | jdbc/TradeDataSource | SELECT t0.ADDRESS, t0.CREDITCARD, ...
0.135ms websphere.datasource.psExecuteQuery | jdbc/TradeDataSource | SELECT t0.ACCOUNTID, t0.BALANCE, ...
...
Dramatically reduce the time it takes to diagnose the source of slow requests
Problem Determination: Request Timing feature
<requestTiming
slowRequestThreshold="10s“
hungRequestThreshold="600s"
includeContextInfo="true"
sampleRate="1"
/>
JBoss and Tomcat do not provide similar capability
=> increased admin labor for troubleshooting

42
Problem Determination: Event Logging feature
Track events running in your Liberty applications
• How do I know what events are happening in the server and how long do they take?
• Create log entries for any Servlet or JDBC request, or HTTP get/set attribute operation
• Use with minimum duration setting to watch for slow events in production environments
• Use with all events enabled to show what apps are doing in development and test
• For best performance use binary logging of Liberty and increase sample rate to >1
Servlets:
[6/18/14 16:21:35:761 IST] 0000002a EventLoggingProbeExtension.class I BEGIN
requestID=AAADvUHkFwy-AAAAAAAAAAD # type=websphere.servlet.service # contextInfo=com.ibm.ws.request.timing.TestJDBC
[6/18/14 16:22:04:643 IST] 0000002a EventLoggingProbeExtension.class I END
requestID=AAADvUHkFwy-AAAAAAAAAAD # type=websphere.servlet.service # contextInfo=com.ibm.ws.request.timing.TestJDBC
# duration=2.614ms
JDBC requests:
[6/18/14 16:21:43:727 IST] 0000002a EventLoggingProbeExtension.class I BEGIN
requestID=AAADvUHkFwy-AAAAAAAAAAD # type=websphere.datasource.executeUpdate # contextInfo=jdbc/exampleDS | create
table cities (name varchar(50) not null, population int, county varchar(30))
[6/18/14 16:21:44:200 IST] 0000002a EventLoggingProbeExtension.class I END
requestID=AAADvUHkFwy-AAAAAAAAAAD # type=websphere.datasource.executeUpdate # contextInfo=jdbc/exampleDS | create
table cities (name varchar(50) not null, population int, county varchar(30)) # duration=0.231ms
Know what's happening in your applications Servlet events include
servlet name, path info, query string
JDBC events include
datasource and SQL
JBoss and Tomcat do not provide similar capability
=> increased admin labor for troubleshooting

43
Previews of technologies that allow you to write engaging and responsive
enterprise business applications
Integrated Analytics
 Ease of Problem
Determination to ensure high
availability of your app-
infrastructure
Intelligent Management
 Additional scenarios for auto
scaling
Liberty server scaling: provisioning
of new server instances to
available host machines
WebRTC
 Write rich, real time multimedia apps
(voice and video) on web without
requiring plug-ins, downloads or
installs.
Strong industry support.
Enables contextual communications!
Java Batch++
 Ease of managing jobs through GUI
tools & support for industry leading
Enterprise Schedulers
e.g. Nightly credit-card processing, bank
reconciliation statements, payroll….
Also available as a Docker container: https://registry.hub.docker.com/_/websphere-liberty/
What’s in WAS Liberty Beta

44
Documentation –
order of magnitude difference in quality
InfoCenter – world class, up to date
Redbooks – unique and comprehensive
developerWorks - implementation tips
ISA – electronic support search tool
3rd party – sites, blogs, etc.
User forums – self help
JBoss docs – limited and inconsistent, lags in time
JBoss wikis – lots of old confusing info
User forums – no longer actively monitored by
developers
Similar quality issues with Tomcat documentation

45
WAS deployment options
On-Premises Public IaaS Public PaaS
Do It Yourself
Business as usual
(can use with IBM
UrbanCode Deploy,
Chef, Puppet, etc.)
BYOL or pay by
the hour on
SoftLayer, Azure,
Amazon EC2
(can use with
IBM UrbanCode
Deploy, Chef,
Puppet)
Liberty Buildpack for
3rd party PaaS
(Cloud Foundry,
OpenShift)
PureApplication
System
PureApplication
System appliance,
or
PureApplication
Software (BYOH)
n/a
PureApp System on
the SoftLayer or 3rd
party cloud
BlueMix BlueMix Local n/a
BlueMix Shared, or
BlueMix Dedicated
or WaaS

46
Cloud Integration
Build hybrid environments. Connect to on-
premises systems of record plus other public and
private clouds. Expose your own APIs to your
developers.
Extend SaaS Apps
CloudFoundry based PaaS from IBM
Run Your Apps
The developer can chose any language runtime or
bring their own. Just upload your code and go.
DevOps
Development, testing, monitoring, deployment
and logging tools allow the developer to run the
entire application
APIs and Services
A catalog of open source, IBM and third party
APIs services allow a developer to stitch
together an application in minutes.
Drop in SaaS App SDKs and extend to new use
cases (e.g,. Mobile, Analytics, Web)
IBM BlueMix

47
Easily deploy, manage and move enterprise applications
without change across Hybrid clouds
• New support for Docker and Chef with
Patterns for 10x faster deployments and
scaling, workload portability and access
to pre-built applications
• Enhanced security and performance for
data and application access across
hybrid environments
• New support for bring your own
hardware and enhanced support for off-
premises cloud environments to
seamlessly deploy and manage
enterprise applications without changes
(with PureApplication Software)
PureApplication
Appliance SoftLayer BYOH
IBM PureApplication Platform

48
A hybrid cloud application platform
for cloud enabling applications and middleware
with enterprise-grade qualities of service
• WAS
• DB2
• Oracle
• MQ
• IIB
Automated elasticity
Multi-site deployment
High availability
Disaster recovery
Monitoring
License management
Intelligent placement
Centralized logging
Security
Over 200 patterns including:
• Portal
• BPM
• Cognos
• DataPower
• MobileFirst
+ any Red Hat/Windows/AIX software
Seamlessly deploy & move workloads
between on & off-premises without
change:
• PureApplication System
• PureApplication Service
• PureApplication Software

49
IBM Pattern Engine Virtual Application Builder
Drag assets onto the
canvas to define
application and
related resources
Define cross-component links and add
policies; respond to warning messages
to build well-formed applications
Specify configuration
details for components,
policies, and links
These patterns can run on-premise or on the IBM SoftLayer cloud

50
IBM PureApplication System business value
9612 hrs
Deployment
Change Management
Security Management
Asset Management
Incident/capacity Mgmt
0
10000
5000
Do It Yourself PureApplication System Pre-integrated
Competitor
Coalition
Competitor
5815 hrs
153%More
4843 hrs
110%More
Labor Hours Spent*
2302 hrs
*Note: Coalition competitor used 9 competitor blades (144 cores). Pre-Integrated competitor used 18 pre-integrated nodes (288 cores). IBM PureApplication System used 3 nodes (96 cores). Each system has the
capacity to run 72 workloads where each workload can sustain a peak throughput of 1720 page elements per second.
The labor savings and assumptions herein are estimates based on a labor model that uses data obtained on the percentage of time customers spend on certain IT life cycle tasks. It is not a benchmark. As such,
actual customer results will vary based on customer applications, differences in stack deployed and other systems variations as well as actual configuration, applications, specific queries and other variables in a
production environment.
76%Savings
How does PureApplication System do this?
- pre-integrated management
- patterns of expertise

51
WebSphere Full Profile secure system administration
Configurator
All Monitor rights plus can make
configuration changes
Operator
All Monitor rights plus can
make runtime changes
Monitor
Can monitor configuration and runtime state
but cannot change them
Administrator
Can monitor and change configuration and runtime state
Operator + Configurator + Monitor rights
Deployer
All Configurator plus Operator
rights plus can deploy projects
AdminSecurityManager
Allows for assignment of user roles and
other security related tasks
JBoss has limited support for role separation (80%)
JBoss does not support resource separation

52
Fine-grained Administrative Security
Key Features:
• Users can be defined with
administrative roles on specific
resources:
• Cells, node groups, nodes,
clusters, servers, and
applications
• Administrative Console will be
filtered by user’s administrative role
• User cannot access any other
resources outside assigned
resources
Isolate administrators from each other and according to access levels to improve security
and governance
JBoss and Tomcat do not provide similar capabilities

53
“ ”
Before… … and after
Invention of “Autopilot”Airplane controls circa 1940
Home grown wsadmin scripts or
“human eyes and hands”
WAS ND
Intelligent
Management

54
WebSphere Intelligent Management
Intelligent
Routing and SLA
Enforcement
Application
Edition
Management
Better TCO through management efficiency and performance, Intelligent
Management delivers the ability to sense and respond quickly to changes
Up to
45%
less hardware
Source: Based on 60+ Operations Optimization Value Assessments done to date by IBM for real customers
Cost reductions are compared to traditional WAS ND deployment
Server Health
Management
SLA based
Dynamic
Clustering
Up to
90%
fewer outages
Up to
60%
less administration
Up to
45%
less software

55
Applications can be upgraded or downgraded without incurring
outages or requiring additional hardware and license costs
Validation
Mode
Rollout
Policies
Concurrent
Activation
Application Edition Management
• Upgrade Applications without interruption to end users
• Concurrently run multiple editions of an application
• Automatically route users to a specific application
• Multiple editions can be activated for extended periods of time
• Rollout policies to switch from one edition to another without service loss
• Easily update OS or WebSphere without incurring down time
• Easy-to-use edition control center in admin console
• Full scripting support
StockTrading 1.0
StockTrading 2.0
StockTrading 3.0

56
WAS
ND
Liberty
WAS
ND
classic
Oracle
WLS
EE
JBoss
EAP
Apache
Tomcat
Hot application update with interruption Yes Yes Yes Yes no
Sequential interruption free update of a compatible app no Yes Yes Yes no
Sequential interruption free update of an incompatible app no Yes Yes no no
Atomic update of an app no Yes no no no
Previous app editions can be activated from history no Yes no no no
Run 2 editions of the same app concurrently no Yes Yes no no
Run app in validation mode while the other edition is running no Yes no no no
Run 3 or more editions of the same app concurrently no no no no no
Application deployment in a clustered environment
• Out of the box function of the products in a multi-server environment is shown
• Most of this function can be achieved with all of the servers with enough work put
into a home grown script or automation framework

57
Sense and respond to problems before end users suffer an outage
Comprehensive
Health Policies
Customizable
Health Conditions
Customizable
Health Actions
57
Health Management
• Automatically detect and handle application health problems
• Without requiring administrator time, expertise, or
intervention
• Intelligently handle health issues in a way that will maintain
continuous availability
• Each health policy consists of a condition, one or
more actions, and a target set of processes
• Includes health policies for common application problems
• Customizable health conditions and health actions
JBoss does not provide similar administrative capabilities

58
Helps mitigate common health problems before outages occur
Health Conditions
• Excessive request timeouts: % of timed out requests
• Excessive response time: average response time
• Excessive garbage collection: % of time spent in GCs
• Excessive memory: % of maximum JVM heap size
• Age-based: amount of time server has been running
• Memory leak: JVM heap size after garbage collection
• Storm drain: significant drop in response time
• Workload: total number of requests
• Health policies can be defined for common
server health conditions
• When a health policy's condition is true,
corrective action execute automatically
or require approval
• Notify administrator (send email or SNMP trap)
• Capture diagnostics (generate heap dump, java core)
• Restart server
• Excessive response time means you are
monitoring what matters most: your customer's
experience!
• Application server restarts are done in a way
that prevent outages and service policy
violations
• Each health policy can be in supervise or
automatic mode. Supervise mode is like training
wheels to allow you to verify that a health policy
does what you want before
making it automatic.
Health Management – Health Policies

59
Flexibility to determine what an “unhealthy” condition is…
 Custom expressions can be built which use metrics from:
• The On Demand Router, URI return codes
• PMI metrics, MBean operations and attributes
• Examples: hung thread detection, DB connection pool exhaustion or slow down
 Complex boolean expressions using a mix of operands is supported (AND, OR, NOT)
Health Management – Custom Health Conditions

60
Provides flexibility by allowing the definition of custom actions
allowing administrators to define an action plan to be carried out
when the unhealthy situation detected.
Health Management – Custom Health Actions

61
• Easily allows an administrator to specify the relative importance of applications
and optionally a response time goal. WebSphere then manages your applications
according to this policy.
• Service policies are used to
define application service level
goals
• Allow workloads to be classified,
prioritized and intelligently routed
• Enables application performance
monitoring
• Resource adjustments are made
if needed to consistently achieve
service policies
Service Policies define the relative importance
and response time goals of application services;
defined in terms the end user result the
customer wishes to achieve
What is a Service Policy?

62
WAS
ND
classic1
WAS
ND
Liberty2
WAS ND +
Data Power
AO8
Oracle
WLS
EE
Traffic
Director &
Exalogic7
JBoss
EAP +
EWP
Apache
HTTP +
Tomcat
High performance some some Yes some some some some
WLM across static app server clusters Yes Yes Yes Yes Yes Yes Yes
SSL termination and HTTP compression Yes Yes Yes Yes Yes Yes Yes
Auto updates of configuration when cluster or app change Yes Yes Yes Yes Yes Yes no9
JVM maintenance mode Yes Yes Yes Yes Yes no no
Node (or host) maintenance mode Yes Yes Yes Yes Yes no no
Application edition-aware routing Yes no Yes Yes Yes no no
Dynamic clusters auto-grow or shrink based on workload Yes Yes Yes no no no no
Health policy support Yes Yes Yes no no no no
Auto-adjusts server weights based on resource use Yes no Yes no Yes no no
Traffic shaping and SLA enforcement for HTTP Yes no Yes no Yes no no
CPU and heap overload protection Yes no Yes no no no no
Support traffic shaping for 3rd party servers Yes no Yes no no no no
Custom rules for request routing (URI, IP address, etc.) Yes no Yes no Yes no no
Request rate limiting no no Yes no some no no
Number of client connections limiting no no Yes no Yes no no
Content based routing no no Yes no some no no
Protect against XML and SQL injection attacks no no Yes no no no no
XML processing (parsing, transformation, validation, etc.) no no Yes no no no no
Custom advisors no no Yes no no no no
DMZ ready Yes5 Yes Yes no no Yes Yes
Static file serving and in-memory and disk page caching Yes3 Yes3 no no4 Yes Yes Yes
Replace hardware based load balancer(s) Yes6 Yes6 Yes no some no no
Traffic shaping and SLA enforcement for IIOP and JMS Yes no no no no no no
HTTP session rebalancing Yes no no no no no no
Intelligent routing

63
Maximum heap
utilization
protects against
OutOfMemory
exceptions
Maximum CPU
utilization
protects
against various
failures which
occur when
CPU is
consumed
Rejects excess
traffic without
affinity when
overload occurs
Overload Protection

64
WAS
ND
classic
WAS
ND
Liberty
Oracle
WLS
EE
JBoss
EAP
Apache
Tomcat
Static clusters (pre-provisioned) Yes Yes Yes Yes Yes
Manually add or remove servers to/from a running cluster Yes Yes Yes Yes Yes
Centralized management of cluster members Yes Yes Yes Yes no
Dynamically creates/starts/stops servers when load changes Yes Yes no no no
Provisions new app servers to hosts when workload increases Yes Yes no no no
Scaling policy allows for min and max number of servers Yes Yes no no no
Scaling policy based on CPU, heap or memory use Yes Yes no no no
Scaling based on service policies (URL + response time, etc.) Yes no no no no
Applications have relative priorities when servers are allocated Yes no no no no
Auto vertical stacking on a node Yes no no no no
Cluster isolation groups Yes no no no no
Lazy application start Yes no no no no
Dynamic clustering (auto scaling)

65
Dynamic clustering
WAS ND classic WLS EE
New node is added
to a cell
If node meets the dynamic selection criteria, it
is automatically added to the dynamic cluster
as potential host for the JVM
Static cluster member must be manually
defined for each participating node and
manually added to the static cluster.
Vertical stacking
(VS)
If VS is allowed, JVM process definitions are
automatically created for each node
Cluster members must be manually created
and port conflict resolution must be manually
done for each new JVM
Cluster isolation
Dynamic cluster can belong to different
isolation groups and conflicts are
automatically resolved
Manual work is required to prevent conflicts
between JVMs that must be isolated from
each other
Workload increase
If workload increases for the application, new
members of dynamic cluster are started to
accommodate such increased workload
Manual start of cluster members is required to
accommodate increase in workload
Workload decrease
When workload drops off, members of
dynamic clusters may be stopped if CPU or
memory are required for other workloads.
Lazy application start can be configured
Manual stop of instances is required to free up
resources for other workloads. Application
must always be up and running to accept
workload
Critical load and
resource shortage
When overall workload is greater than the
system can handle, service policies are
enforced such that more important
applications get priority over less important
ones and SLA policies for response times are
met. SLAs can be defined based on a rule set
based on URI, time, user properties, IP, etc.
No provision for prioritization of workload, no
SLAs for applications. Typical solution is to
create duplication by using dedicated hosts
(physical or virtual) for each workload, which
increases admin complexity, hardware and
software cost
Server properties
Server template can be updated and changes
are reflected on all members of dynamic
cluster automatically
Properties must be updated on each member
of the static cluster manually

66
Dynamic clustering
WAS ND JBoss EAP
New node is added
to a cell
If node meets the dynamic selection criteria, it
is automatically added to the dynamic cluster
as potential host for the JVM
Static cluster member must be manually
defined for each participating node and
manually added to the static cluster.
Vertical stacking
(VS)
If VS is allowed, JVM process definitions are
automatically created for each node
Cluster members must be manually created
and port conflict resolution must be manually
done for each new JVM
Cluster isolation
Dynamic cluster can belong to different
isolation groups and conflicts are
automatically resolved
Manual work is required to prevent conflicts
between JVMs that must be isolated from
each other
Workload increase
If workload increases for the application, new
members of dynamic cluster are started to
accommodate such increased workload
Manual start of cluster members is required to
accommodate increase in workload
Workload decrease
When workload drops off, members of
dynamic clusters may be stopped if CPU or
memory are required for other workloads.
Lazy application start can be configured
Manual stop of instances is required to free up
resources for other workloads. Application
must always be up and running to accept
workload
Critical load and
resource shortage
When overall workload is greater than the
system can handle, service policies are
enforced such that more important
applications get priority over less important
ones and SLA policies for response times are
met. SLAs can be defined based on a rule set
based on URI, time, user properties, IP, etc.
No provision for prioritization of workload, no
SLAs for applications. Typical solution is to
create duplication by using dedicated hosts
(physical or virtual) for each workload, which
increases admin complexity, hardware and
software cost
Server properties
Server template can be updated and changes
are reflected on all members of dynamic
cluster automatically
Properties must be updated on each member
of the static cluster manually

67
The On Demand Router applies sophisticated classification and flow control algorithms to
intelligently manage workload
WebSphere On Demand Router
(ODR)
Classification
Prioritization and
Flow Control Routing and Load
Balancing
Placement
Executions
Node 2
Node 3
Node 4
Node 5
Placement
DecisionsWebSphere Decision
Makers
Application Demand
Resource State
Routing, Health and
Service Policies
Node 1
Intelligent Management Scenario

68
68
Proactively provision and start or stop application servers based
on workload demands to meet Service Level Agreements
Dynamic Clustering
 Associate service policies with your applications
• Let WebSphere manage to the service goals
 Programmatically respond to spikes in demand
• Add or reduce application server instances as appropriate
 Automatically recover from infrastructure problems
 Includes automatic start and stop of cluster members based on load for
MQ-driven applications
 Decrease administrative overhead required to monitor and diagnose
performance issues

69
Lazy Application Start
• Dynamic Clusters support a min and max number
of running cluster members
• If the “stop all instances” option is enabled, the min
is 0, which means:
 The application may not be running in the pool
anywhere
 When a request is received, a cluster member
is started
 When the application goes idle all clusters
members are stopped
 This allows low volume applications to be
available without consuming resources.
 A customizable On Demand Router error page
with meta-refresh provides a user-friendly
customer experience

70
Before Application Edition
Management
1 Stop application servers
2 Uninstall old version of application
3 Install new version of application
4 Replicate application changes to all
nodes
5 Start application servers
Application is unavailable from step 1
through 5
To revert to old version of application,
repeat all steps, reversing “old” and
“new” ... thus another long outage
With Application Edition
Management
1 Install new edition of application
2 Replicate application changes to all
nodes
3 Roll out new edition of application
Application remains available to end
users throughout the update process
To revert to old version of application,
simply rollback the old edition
Eliminate down-time for managed applications

71
Administrative Console - Edition Control Center

72
Memory leak detection & Protection in WAS
WebSphere Application Server V8.5:
• Ability to mitigate memory leak when stopping apps
• Ability to prevent leaks, receive leak warnings and get heap/system
dumps
• MBean to list stopped apps that have memory leaks
72
Reduce possibilities
of memory leak in
your applications
Get enough info.
if leak is detected
to help fix my app
List stopped
apps that have
memory leaks

73
The history of Red Hat and JBoss messaging
JBoss
AS v3
2002 2006
JBossMQ
JBoss
AS v5
JBoss Messaging
2009
JBoss
AS 6
HornetQ
2013
JBoss
XQ
ActiveMQ
Red Hat MRG2008
Apache Qpid
* - New Red Hat “strategic” messaging is described to be
a REWRITE and a combination of “best ideas” from
Apache Qpid + Red Hat HornetQ + Apache ActiveMQ
2015 NEW* ?
?

74
Platforms support WAS
8.5.5.7
WebLogic
12.2.1
JBoss EAP
6.4
tc Server
X86 Red Hat Enterprise Linux 5, 6, 7 6, 7 6, 7 6, 7
Asianux 3 no no no
Ubuntu 12, 14 no no Dev
Oracle Linux no 6, 7 no no
Mac OS Liberty Dev. no Dev
SuSe Linux ES 10, 11, 12 11 no V11
Windows 7+, 2008+ 2012 2008+ 2008
Solaris 10, 11 11 11 no
RISC Red Hat Enterprise Linux 5, 6, 7 6 7 no
SuSe Linux ES 10, 11, 12 11 no no
Ubuntu 14 no no no
AIX 6, 7 7 no no
IBM I 6, 7 no no no
HP-UX 11i v2, v3 11i v3 11i v3 no
Inspur K-UX (Itanium) 2.1 no no no
Solaris (SPARC) 10, 11 11 11 no
z/Series Red Hat Enterprise Linux 5, 6, 7 6 no no
SuSe Linux ES 10, 11, 12 11 no no
z/OS 1, 2 no no no

75
WAS
8.5.5.7
WLS
12.2.1
Oracle 11g, 12c 11g, 12c
Microsoft SQL 2008, 2012 2008, 2012
Sybase 15 15
IBM DB2 9, 10 9, 10
IBM DB2 for iSeries 6, 7 no
IBM DB2 for z/OS 9, 10, 11 no
IBM Informix DS 11, 12 no
Apache Derby 10.8 no
MySQL no 5 (no XA)
Why do I care? IBM offers more choices and allow to pick the right product for the right
job, which often can reduce the cost of computing.
Database certifications

76
Zero migration by design
• Why upgrade?
• Gain new APIs in a new Java EE level, stay on a
supported version of the Application Server, protect against security vulnerabilities,
gain speed by moving to the faster new version of JDK or newer hardware
• How easy is it to upgrade? Does it cost a fortune?
• Liberty implements Zero Migration by design
• There is no migration needed for Liberty configuration - the same server
configurations can be used with different versions and service levels
• Existing features will not change behavior, new feature ‘versions’ will be added and
will contain all updates and changes
• Tomcat and tc Server require considerable migration effort
• Fresh server install and careful configuration updates and testing are required
• No automated upgrade or migration tool is provided
• JBoss is not backwards compatible and migration is a pain
• Backwards compatibility was broken between JBoss v3.x, v4.x, v5, v6. Each of
these releases have been disruptive and changed many properties and
configuration files, scripting commands, CLI, Admin UI, APIs, etc. Upgrade path for
JBoss is to manually copy configuration files and applications to new installation.
• These issues result in increased administration costs when using JBoss because
of lost productivity related to unnecessary software development.

78
Considerations for moving off WAS
What do you expect to gain from switching? What is the exact criteria to measure
if it was a viable move? Moving applications from WAS to JBoss will involve
significant effort and cost, including:
• Training for operations team
• Migration of operational scripts and documentation
• Validate the operational characteristics of JBoss
• Building missing functions in JBoss
• Purchase additional hardware capacity
• Purchase more licenses for JBoss
• Security
• Additional license costs
• Unreliable messaging in JBoss
• Lack of dynamic clustering and SLA management in JBoss
• There are many other capabilities included with WAS that JBoss does not
provide
See speaker notes for details

79
WebSphere Application Server Migration Toolkit
• No Charge plugins for Eclipse and RAD
Rule sets for multiple source / destination
combinations (e.g. WLS->WAS, etc.):
(a) The tool scans Java source code, JSP files
and deployment descriptors and identifies the
changes required (allows for Java upgrade
also).
(b) The tool scans server configuration files
(looking for Datasources, servers, JMS settings,
etc.) and generates appropriate Liberty or WAS
configuration.
In most cases the toolkit is capable of making
the application changes itself. After the “scan”
and “conversion” are done the toolkit generates
report on the results of the migration and any
manual migration tasks (if required).
• Free migration RedBook and
developerWorks articles on migration
• No Charge Migration Assessment
Workshop for qualified customers
Now easier then ever before to migrate your applications to WebSphere Application Server
(1) Liberty Profile or (2) WAS
v7, v8, v8.5, v8.5.5
WebSphere Migration Toolkit
(Eclipse and RAD plugins)
WAS
5.1 – 8.x
WebLogic
Oracle
OC4J
(OAS)
Tomcat
IBM migration tools and offerings: http://whywebsphere.com/?s=migration
(a) Java, JSP source and DDs
(b) Server configuration objects
JBoss
new

80
The Migration tool in action…
Analysis Type
Rule Categories
Rule
Result Options
Rule Results
Analysis
History
Analysis Results
Help Contents

81

82
and Tomcat plus Liberty Technology Preview plugins
From
To
Liberty
Config
Migration
Liberty Application
Migration
WAS
Config
Migration
WAS
Application
Migration
JBoss 4.X – 5.x 
Liberty 8.5.5 or
CloudFoundry /
Bluemix
Java EE5 and prior
versions 
Liberty 8.5.5 or
CloudFoundry / Bluemix
4.X – 5.x 
7.0 - 8.5.5
Java EE5 and prior
versions 
7.0 – 8.5.5
Tomcat 7.X 
Liberty 8.5.5
6.0 or 7.0 
Liberty 8.5.5 or
N/A 6.0 or 7.0 
7.0 - 8.5.5
WebLogic 6.X – 11.x 
Liberty 8.5.5 or
CloudFoundry /
Bluemix
Java EE5 and prior
versions 
Liberty 8.5.5 or
6.X – 11.x 
7.0 - 8.5.5
Java EE5 and prior
versions 
7.0 – 8.5.5
OAS N/A Java EE5 and prior
versions 
Liberty 8.5.5 or
N/A Java EE5 and prior
versions 
7.0 – 8.5.5
WAS N/A 7.0 - 8.5.5 
Liberty 8.5.5 or
N/A 5.1 – 8.x 
7.0 - 8.5.5
Java (JDK) N/A 1.4, 5.0, 6.0 
6.0 or 7.0
N/A 1.4, 5.0, 6.0 
6.0 or 7.0

83
Capitalizing on Intelligent App Server Management
Base Application Server
• Programming Model
• QoS
• Security
•Administration
Server
Server
Server
Server
Server
Server
Server
Server
Job Manager
• Control multiple endpoints
• Remote management
• Loose Coupling
Network Deployment Cell
•Administration
•Clustering
•Workload Management
WebSphere
Application Server
Server Server
Server
Server
Server
Server
Server
Admin
Agent
Admin
Agent
Admin
Agent
Deployment
Mgr
Deployment
Mgr
Read more details here: http://smarterquestions.org/2012/01/comparison-of-automation-tools-for-large-scale-websphere-weblogic-and-jboss-topologies
Low cost administration of massive remote or local installations

85
JBoss Operations Network
Number of stability
issues and bugs
remain open. Tool is
not mature enough
to be used in full
production

86
Administration and Configuration with WebSphere is
Easier: Cluster Configuration
WAS ND 8.5 has a UI tc Server 2.8.1:
Hand-edited XML
Hand-editing XML files leaves no audit trail for changes!
Cluster Configuration

87
Management options
Pros Cons
Manual editing of
files
• Easy to understand
• Best for development use
• Not reliable – user can make typos
and break configuration, leading to
costly outages
• Time consuming when managing
more than 1 server
• No auditing, limited security
• Not recommended for production
Administrative GUI
• Easy to understand
• Auditing and security provided
• Configuration consistency checks
• Some group operations supported
• Best for development use
• Often requires repetition of commands
to be applied to multiple servers
• Despite configuration consistency
checks and input validation, manual
keystrokes and mouse clicks may lead
to errors and downtime in production
• Not recommended for production
Command Line
Script
• Repeatable and predictable (no user input is
needed, no typos, no wrong mouse clicks)
• Can automate management of large
configurations by iterating over resource lists
(no need to manually repeat steps)
• Best for production use
• Can be difficult to learn and master
• High overhead for development use

88
Why choose Liberty over JBoss EAP?
• Reduced development costs and more agile smaller runtime
• Liberty offers full Java EE 7 certified runtime, JBoss EAP is Java EE 6
• Integrated security options
• Liberty provides multiple security options and avoids security risks of open source (like
Heartbleed bug)
• Lower hosting costs than JBoss EAP
• Higher throughput on each server means fewer resources are needed to support the
same workload on Liberty (Liberty throughput is 30% higher than JBoss on native OS
and up to 2.1x better on 4 Docker Container topology)
• Lower license and support costs with Liberty Core and Base
• Operational capabilities and QOS
• Cluster auto scaling, soon new Intelligent Mgmt features, dynamic configuration updates
• Painless V2V migration
• Put all your development effort on new function, not migration
• Application fidelity between Liberty and WAS
• Fast low cost startup and easy growth on IBM hosted cloud
• Bluemix public for instant environments
• Seamless growth to Bluemix dedicated/local and IaaS Clouds
• World class IBM support
• Liberty Core edition is free to embed and ship in ISV products (optional $ support)
• IBM ships & supports entire stack – from JVM and app server, to HTTP, LDAP, etc.
• Blue Diamond support option in North America for enhanced data privacy

89
Why choose Liberty over Tomcat and TomEE?
• Faster development cycle / reduced development costs
• Users who move application development from Tomcat to Liberty report 20-25%
increase in development efficiency
• Liberty offers full Java EE 7 certified runtime, Tomcat is only JSP/Servlet engine
• Integrated security options
• Liberty provides multiple security options and avoids security risks of open source (like
Heartbleed bug)
• Lower hosting costs than Tomcat
• Higher throughput on each server means fewer resources are needed to support the
same workload on Liberty (Liberty throughput is 34% higher than Tomcat on native OS
and 2.4x better than Tomcat on 4 Docker Container topology)
• Operational capabilities and QOS
• Admin center, cluster auto scaling, dynamic routing updates, soon new Intelligent Mgmt
features, transaction management, JMS messaging, dynamic configuration updates
• Painless V2V migration
• Put all your development effort on new function, not migration
• Application fidelity between Liberty and WAS
• Fast low cost startup and easy growth on IBM hosted cloud
• Bluemix public for instant environments
• Seamless growth to Bluemix dedicated/local and IaaS Clouds
• World class IBM support
• Liberty Core edition is free to embed and ship in ISV products (optional $ support)
• IBM ships & supports entire stack – from JVM and app server, to HTTP, LDAP, etc.
• Blue Diamond support option in North America for enhanced data privacy

90
Financial Considerations
TCA vs. TCO
( )f=

91
Cloud support?
Standards support and
programming model?
Monitoring and
diagnostic tools?Management and
administration?
High availability
and reliability?Performance and scalability?
User and administrative
security?
Minimize License and support cost (TCA)?
OS and DB support?
Documentation
and best
practices?
PaaS?
Minimize TCO
Time to market?

92
Average cost of downtime per industry
Industry segment Cost per Hour
(Millions)
Energy $ 2.8
Telecommunications $ 2.1
Manufacturing $ 1.6
Financial $ 1.5
Information Technology $ 1.4
Insurance $ 1.2
Retail $ 1.1
Pharmaceuticals $ 1.1
Banking $ 1.0
Consumer Products $ 0.8
Chemicals $ 0.7
Transportation $ 0.7
Sources: ITG Value Proposition for Siebel Enterprise Applications, Business case for IBM System z & Robert Frances Group
&*^$#@ ???
Zzzzzzz….

93
<10%• Software license &
subscription costs1
• Hardware and networking costs
• Downtime costs (planned and unplanned)
• Upgrades cost
• SLA penalties
• Deployment cost
• Operational support cost (day to day operations)
• Performance costs
• Cost of selection of the vendor software
• Requirements analysis cost
• Developer, admin and end-user training cost
• Application design and development costs
• Cost of integration with other systems
• Quality, user acceptance and other testing costs
• Application enhancements and bug fixes cost
• Replacement costs
• Cost of other risks (including security breaches)
90%
(1) Source: http://bit.ly/1yH5oKZ

94
Free like in beer
•NO CHARGE WebSphere Developer Tools for Eclipse
•NO CHARGE WAS for Developers & Liberty Profile
Available at no charge for the developer desktop/laptop – free license +
free support for those who have production licenses, and optional fee
based support for those who don’t
•NO CHARGE production runtime – Liberty Core for ISVs
ISV’s customers can run the app on Liberty Core free of charge without
support
•NO CHARGE production – Liberty for up to 1GB on BlueMix
Liberty instance for test or production running non-stop
•NO CHARGE production runtime – Liberty for up to 2 GB
Any number of instances, so long as sum total Java heap is <=2GB

95
Flexible licensing options to suit customer needs
• JBoss, tc Server do not have socket, or per user pricing
• tc Server, WebLogic and JBoss do not have On-Demand per day pricing
• Consider retail chain with 100s of locations. WAS user based license for Liberty
Core or Express can be orders of magnitude less than JBoss or tc Server
See additional notes about these pricing options here: http://whywebsphere.com/2013/09/26/software-costs
Core
(PVU)
Socket 20 users
Unlimited
license
1 hour
(cloud)
1 day
(Power)
1 month
(PVU)
1 month
(socket)
1 year
(PVU)
WAS Liberty Core 28.25$ n/a 709$ BYOL 25.50$ 1.18$ n/a 11.40$
WAS Express 28.25$ n/a 709$ BYOL 25.42$ 1.18$ n/a 11.40$
WAS Base 57.00$ 14,500$ n/a 0.53$ 51.09$ 2.38$ 604$ 22.80$
WAS ND 214.00$ n/a n/a 1.11$ 191.67$ 8.90$ n/a 88.25$
Pay as you go
Contact
IBM
Perpetual licenses

96
License cost calculator, blog and video
• WAS vs. WLS: http://bit.ly/1JZFy8V
• WAS vs. JBoss: http://bit.ly/22aOKkz

97
48 53 58
72
58
43 38
48
62
96
384
144
Example of the use of monthly term license
Workload distribution example over calendar year (hypothetical)
Servers are 2 sockets, 12 cores each
Numberofcores
JBoss subscription licenses: 384=16*24
JBoss 5 year cost = $1.44M
WAS perpetual licenses: 8 sockets
WAS monthly licenses: 336
IBM 5 year cost = $283K
For applications that have
uneven workloads over the year
the cost of WAS could be
minimized by purchasing “pay
as you go” licenses for peak
periods

98
48 53 58
72
58
43 38
48
62
96
384
144
Example of the use of monthly term license
Workload distribution example over calendar year (hypothetical)
Servers are 2 sockets, 12 cores each
Numberofcores
Oracle perpetual licenses: 384=16*24
Oracle 5 year cost = $3.44M
WAS perpetual licenses: 8 sockets
WAS monthly licenses: 336
IBM 5 year cost = $283K
For applications that have
uneven workloads over the year
the cost of WAS could be
minimized by purchasing “pay
as you go” licenses for peak
periods

99
WAS license + support cost over 5 and 10 years is lower
And it gets better…
See additional notes about these pricing options here: http://whywebsphere.com/2013/09/26/software-costs
Without required components With LDAP, JDK and HTTP
5 years 10 years 5 years 10 years
WAS JBoss EAP WAS JBoss EAP WAS JBoss EAP WAS JBoss EAP
4 1 4 50 x86 $61,560 $90,000 $95,760 $180,000 $61,560 $211,875 $95,760 $423,750
4 1 6 50 x86 $78,300 $90,000 $121,800 $180,000 $78,300 $253,125 $121,800 $506,250
4 1 8 50 x86 $78,300 $135,000 $121,800 $270,000 $78,300 $335,625 $121,800 $671,250
4 1 12 50 x86 $78,300 $180,000 $121,800 $360,000 $78,300 $468,750 $121,800 $937,500
4 1 16 50 x86 $78,300 $270,000 $121,800 $540,000 $78,300 $637,500 $121,800 $1,275,000
4 2 4 70 x86 $156,600 $135,000 $243,600 $270,000 $156,600 $335,625 $243,600 $671,250
4 2 6 70 x86 $156,600 $180,000 $243,600 $360,000 $156,600 $468,750 $243,600 $937,500
4 2 8 70 x86 $156,600 $270,000 $243,600 $540,000 $156,600 $637,500 $243,600 $1,275,000
4 2 10 70 x86 $156,600 $315,000 $243,600 $630,000 $156,600 $766,875 $243,600 $1,533,750
4 2 12 70 x86 $156,600 $360,000 $243,600 $720,000 $156,600 $890,625 $243,600 $1,781,250
4 2 14 70 x86 $156,600 $450,000 $243,600 $900,000 $156,600 $1,102,500 $243,600 $2,205,000
4 2 16 70 x86 $156,600 $495,000 $243,600 $990,000 $156,600 $1,226,250 $243,600 $2,452,500
4 2 18 70 x86 $156,600 $540,000 $243,600 $1,080,000 $156,600 $1,359,375 $243,600 $2,718,750
4 4 6 100 x86 $313,200 $360,000 $487,200 $720,000 $313,200 $890,625 $487,200 $1,781,250
4 4 8 100 x86 $313,200 $495,000 $487,200 $990,000 $313,200 $1,226,250 $487,200 $2,452,500
4 4 10 100 x86 $313,200 $585,000 $487,200 $1,170,000 $313,200 $1,479,375 $487,200 $2,958,750
4 4 12 100 x86 $313,200 $720,000 $487,200 $1,440,000 $313,200 $1,781,250 $487,200 $3,562,500
4 4 14 100 x86 $313,200 $855,000 $487,200 $1,710,000 $313,200 $2,116,875 $487,200 $4,233,750
4 4 16 100 x86 $313,200 $945,000 $487,200 $1,890,000 $313,200 $2,370,000 $487,200 $4,740,000
4 4 18 100 x86 $313,200 $1,080,000 $487,200 $2,160,000 $313,200 $2,671,875 $487,200 $5,343,750
CPUtype
#ofphysicalservers
#socketsperserver
#corespersocket
IBMPVUrating

100
License cost comparison of additional components for App
Server
WAS WAS ND JBoss EAP
Management and monitoring Included Included Included (in “managed” bundles)
JON configuration DBMS n/a n/a $6,900 / CPU / year (PostgreSQL)
Hardware for the JON database n/a n/a ~ $15,000 + support (3rd party)
Load Balancer Extra $ Included ~ $20,000 / device + support (3rd party)
Dynamic content caching proxy Extra $ Included $2,500 / 16 cores / year (JBoss EWS)
Page fragment & POJO caching Included Included ~ $1,000 / server / year (3rd party)
HTTPSession persistence DBMS Included Included $6,900 / CPU / year (PostgreSQL)
LDAP Included Included $9,000 / server / year (3rd party)
JDK Included Included OpenJDK is supported on RHEL
$5,000 / core (Oracle JDK)
Troubleshooting tools Included Included $?,000 / year (3rd party)
HTTP Server Included Included $2,500 / 16 cores / year (JBoss EWS)
App Server Hardware $X $X $X + 30% (due to lower performance)

101
Support policy for IBM vs. Red Hat
• Fee based 24x7 Production support
• all cores in production must be licensed
• Fee based Non-Production
• WAS, MQ, JBoss A-MQ, JBoss EAP must be licensed for non-production
• See Red Hat subscription guide (details here)
• Free Disaster Recovery
• Cold and Warm Standby as well as DR servers are provided at no extra charge
with IBM and Red Hat contacts (this is very unlike Oracle)
• Free Development support
• MQ, WAS for Developers (including Liberty) – unlimited developer support with
at least one production license
• JBoss A-MQ, JBoss EAP, etc. – up to 25 developers for each 16 cores of
subscription for production support (see details here)
• Number of production support contacts
• IBM: unlimited
• Red Hat: depends on the number of cores licensed: 2 contacts up to 32 cores,
4 contacts up to 64 cores, etc. up to 12 contacts for 192 cores (more details)

102
Forrester case study 1: The Total Economic Impact To IBM
WAS Migrating From An Open Source Environment
 Case study of a US
Government Agency
 Migration of the JBoss
production system to
WebSphere Application
Server yielded 44% three-
year risk-adjusted ROI
with payback period of 2
years

103
Forrester case study 2: The Total Economic Impact of IBM
WebSphere Application Server
• Case study of a US based Fortune 100
company
• Migration of the JBoss production system
to WebSphere Application Server yielded
42% ROI and payback of 1.4 years
• Primary benefits of migration
• Improved administration
• Greater application performance
• Higher application availability
• Reduced support costs
• Improved development
productivity

104
Independent industry studies on software costs
“Software maintenance and evolution is a considerably understudied area while taking into account its cost
effects… Although there has not been much empirical research on this particular area, the magnitude of
the maintenance cost effects is clearly identifiable. The relative cost for maintaining software and
managing its evolution now represents more than 90% of its total cost. This is referred to as legacy
crisis by Seacord et al. (2003).”
Various studies on this subject are described in the table below:
Year Proportion of software
maintenance costs Definition Reference
2010 >81%
Implementation, testing, management, tuning and
other costs (all, but for procurement)
Forrester Research (2010)
2000 >90%
Software cost devoted to system maintenance
& evolution / total software costs
Erlikh (2000)
1993 75%
Software maintenance / information system budget
(in Fortune 1000 companies)
Eastwood (1993)
1990 >90%
Software cost devoted to system maintenance
& evolution / total software costs
Moad (1990)
1990 60-70%
Software maintenance / total management
information systems (MIS) operating budgets
Huff (1990)
1988 60-70%
Software maintenance / total management
information systems (MIS) operating budgets
Port (1988)
1984 65-75%
Effort spent on software maintenance /
total available software engineering effort.
McKee (1984)
1981 >50%
Staff time spent on maintenance / total time
(in 487 organizations)
Lientz & Swanson (1981)
1979 67% Maintenance costs / total software costs Zelkowitz et al. (1979)
Source: http://archive.is/oBlIr

105
Red Hat JBoss support metric

106
TCO study: WAS ND 8.5 vs. JBoss EAP v6
Conclusion: JBoss is 35% more
expensive over 5 years vs. WAS ND
Source: Based on the study by Prolifics, December 2012
TCO Category IBM Red Hat
RedHat as
% of IBM
Hardware $ 2,060,934 $ 3,114,308 151%
Training $ 84,375 $ 171,998 204%
Software License $ 2,623,920 $ - 0%
Software Support $ 2,008,815 $ 1,821,316 91%
Application Management $ 759,492 $ 2,570,500 338%
Infrastructure Management $ 1,533,834 $ 2,301,566 150%
Risk and Downtime $ - $ 2,268,548 n/a
Total $ 9,071,370 $ 12,248,235 135%

107
Cost of Development vs. the rest of the TCO
 Financial Services company in the US (Multi-billion dollar business)
• 5 years time horizon
• 30 JEE applications
• about 6 new apps built per year
• 110 developers, approximate *loaded* cost is $145,000 per year
• 24 servers (16 cores each, x86), approximate cost $23,000 plus support, power,
cooling, etc.
• Lesson learned
• The cost of the IBM software is 4% of the TCO
TCO Category 5 year cost
Hardware $ 1,853,048
Training $ 1,115,625
Software Licenses $ 2,208,960
Software Support $ 1,721,199
Application Management $ 736,596
Infrastructure Management $ 1,484,757
Development $ 78,817,500
Total $ 87,937,684

108
WAS TCO advantage over Pivotal tc Server
• License and support cost (for WAS
Express and Base)
• Performance
• Security
• Lower downtime
• Monitoring and troubleshooting
• Server health monitoring
• SLA enforcement at runtime
• Dynamic clustering
• More complete Java EE compliance
Application Server Total Cost of Ownership over 5 years (no JDK)

109
WAS tc Server
VMware as
% of IBM
Hardware w/ support and power $ 532,571 $ 712,822 134%
Software w/ support $ 448,383 $ 893,314 199%
Application Management $ 200,624 $ 397,751 198%
Infrastructure Management $ 442,051 $ 524,402 119%
Risk and Downtime $ - $ 340,655 n/a
Training $ 93,750 $ 101,175 108%
Development $ - $ - 0%
Total $ 1,717,379 $ 2,970,120 173%
TCO study results: WAS vs tc Server
WAS tc Server
VMware as
% of IBM
Hardware w/ support and power $ 532,571 $ 712,822 134%
Software w/ support $ 448,383 $ 442,264 99%
Application Management $ 200,624 $ 397,751 198%
Infrastructure Management $ 442,051 $ 524,402 119%
Risk and Downtime $ - $ 340,655 n/a
Training $ 93,750 $ 101,175 108%
Development $ - $ - 0%
Total $ 1,717,379 $ 2,519,070 147%
1. Without JDK support
(free OpenJDK)
2. With JDK support
(provided by Oracle)

110
Oracle software licensing does not permit soft partitioning
Logical/soft
partition with
WAS ND on
2 cores
Logical/soft
partition without
WAS on 6 cores
WebSphere AS ND is licensed for 2 cores
License & support cost for 5 years= $53,928
WebLogic Server Enterprise is licensed for 8 cores.
License & support cost for 5 years = $210,000
VMware image with
WebLogic Server
Enterprise on 2 cores
VMware images without
WebLogic on 6 cores
(one still must pay for
these)
You pay Oracle for all CPUs on a server vs. CPUs that are assigned to the logical VM.
Oracle does not allow the use of soft partitioning as a means to determine or limit the number
of software licenses required for any given server.
Read detailed analysis here: http://bit.ly/1JZFy8V
Example:
Based on publicly available information as of 7/10/2015 comparing Oracle WebLogic Server Enterprise Edition to IBM WebSphere Application Server Network
Deployment. Both include maintenance and support for 5 years. IBM: 70 Processor Value Units per core, Oracle: 0.5 processor multiplier, both are on an x86 server,
2 sockets, quad core each.

111
Virtualization and server partitioning support5
1 - Oracle does not certify nor supports certain 3rd party software hypervisors as shown in the Oracle column (No means no support, Yes means support).
2 - Oracle charges up to full capacity of the servers, regardless of the number of cores used, except for some hypervisors (No means charge for all cores on a server, Yes means charge only for cores assigned to a VM).
3 - Turbocharged cores are not eligible for reduced Oracle pricing on Power7.
4 – Not all configurations of OracleVM and Solaris Containers are eligible for reduced Oracle pricing.
5 - Read more details here: http://whywebsphere.com/2012/02/16/ibm-and-oracle-software-licensing-and-support-in-virtualized-private-cloud-environments/
Support1 Reduced pricing2
IBM Oracle IBM Oracle
VMware Yes no Yes no
IBM z/VM Yes no Yes no
IBM PR/SM Yes no Yes no
IBM PowerVM LPAR Yes Yes Yes Yes/no3
Xen Yes no Yes no
Red Hat KVM Yes no Yes no
Hyper-V Yes Yes Yes no
Xen Yes no Yes no
Oracle VM Yes Yes Yes Yes/no4
Solaris containers Yes Yes Yes Yes/no4

112
Virtualization cost comparison (license + support over 5 years)
Example 1: One x86 server with VMware, 2 sockets, 8 cores total, 2 cores used for the workload
IBM
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years Oracle
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years
WebSphere Application Server ND 2 $26,600 $21,280 $47,880 WebLogic Server Enterprise Edition 8 $100,000 $110,000 $210,000
IBM DB2 Enterprise Edition 2 $58,380 $46,704 $105,084 Oracle DB Enterprise Edition 8 $190,000 $209,000 $399,000
WebSphere eXtreme Scale 2 $21,280 $17,024 $38,304 Oracle Coherence Grid Edition 8 $100,000 $110,000 $210,000
IBM
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years Oracle
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years
IBM
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years Oracle
#corestobe
licensed
License
cost
Support
cost over
5 years
Total license
+ support
over 5 years
List prices are used for cost comparisons. Actual environmental costs and performance characteristics will vary depending on individual client configurations and conditions.
Read more details here: http://smarterquestions.org/2012/02/ibm-and-oracle-software-licensing-and-support-in-virtualized-private-cloud-environments

113
Oracle charges more for backup and disaster recovery
• Both IBM and Oracle charge for the
main cluster and hot backup
• Oracle charges full license cost for
“Warm” backup servers
• IBM does not
• Oracle charges full license cost for
“Cold” backup servers in DR setup
• IBM does not
• Oracle charges for “Cold” backup
when failover is > 10 days
• IBM does not
Main cluster Warm backup Cold backup Disaster
Recovery
$
No Charge
$ $ $
$
$
$
Hot backup
Example: x86 server, 2 sockets, 8 cores total
IBM
License +
support
Oracle
License +
support
Main cluster WAS ND $191,520 WLS EE $210,000
Hot backup WAS ND $191,520 WLS EE $210,000
Warm backup WAS ND $0 WLS EE $210,000
Cold backup* WAS ND $0 WLS EE $210,000
Disaster recovery WAS ND $0 WLS EE $210,000
Total 5 year cost $383,040 $1,050,000
* - failover to cold backup for more than 10 days in a year
No Charge No Charge

114
Support costs
Support includes
version upgrades
  Support includes version upgrades
First year of support
is at no extra charge
  First year of support costs additional
22% on top of the license cost
Support is 20% of the
license cost
  Support is 22% of the license cost
Support cost is a % of the
PPA entitled price
  Support cost increases at least 4%
per year until reaches list price
Critical fixes are provided even
without current support
  Critical fixes are not provided without
current support
IBM software running on 3rd party
hypervisors is supported
  Oracle software running on 3rd party
hypervisors is not supported*
* - More details on virtualization support: http://smarterquestions.org/2012/02/ibm-and-oracle-software-licensing-and-support-in-virtualized-private-cloud-environments

115
Example: x86 servers, no virtualization, no backup
IBM
cores
License +
support
Oracle
cores
License +
support
WebSphere Application Server ND 12 $287,280 Oracle WebLogic Server Enterprise Edition 12 $315,000
WebSphere Edge Cache (included) 4 $0 Oracle Web Cache (Oracle Web Tier) 4 $21,000
WebSphere Edge WLM (included) 4 $0 3rd party load balancer (hw based) 4 $42,000
DB2 UDB (included) 4 $0 Oracle DB Enterprise (for session replication) 4 $199,500
IBM HTTP Server (included) 8 $0 Oracle HTTP Server (Oracle Web Tier) 8 $42,000
Tivoli Directory (included) 4 $0 Oracle Directory Services 4 $46,200
$287,280 $665,700
WAS ND vs. WLS Enterprise pricing (5 years)
IP Sprayers Caching
Servers
HTTP
servers
JEE servers
LDAP
servers
Oracle: $cost
IBM: $0
Session DB
servers
IBM: $0 IBM: $0
IBM: $cost
IBM: $0
IBM: $0
Oracle: $cost
Oracle: $cost
Oracle: $cost
Oracle: $cost
Oracle: $cost

116
IBM
#ofcores
License +
support
Oracle
Hotcluster
Warmbackup
DR
Coldbackup
Virtualization
#ofcores
License +
support
WebSphere Application Server ND 16 $383,040 Oracle WebLogic Server Enterprise Edition 16 8 8 8 14 54 $1,417,500
WebSphere Edge Cache (included) 4 $0 Oracle Web Cache (Oracle Web Tier) 4 2 2 2 4 14 $73,500
WebSphere Edge WLM (included) 4 $0 3rd party load balancer (hw based) 4 2 2 8 $84,000
DB2 UDB (included) 4 $0 Oracle DB Enterprise (for session replication) 4 0 2 2 3 11 $548,625
IBM HTTP Server (included) 4 $0 Oracle HTTP Server (Oracle Web Tier) 4 2 2 2 4 14 $73,500
Tivoli Directory (included) 4 $0 Oracle Directory Services 4 2 2 2 4 14 $161,700
$383,040 $2,358,825
Example representing an extreme case of failover and
redundancy highlights compounding effect of Oracle terms
• Compounding effect of all the license terms and conditions that Oracle
imposes on customers results in large software license and support costs1
• Higher license costs2, higher support costs, cost of warm backup, cold backup,
DR, no support for virtualization, lower performance per core3
$383K $2.3M
1 - the cost comparison is done over 5 years assuming x86 servers with 70 PVU core rating for IBM and 0.5 core factor for Oracle
2 - List prices are used for cost comparisons. Actual environmental costs and performance characteristics will vary depending on individual client configurations and conditions.
3 - performance metrics derived from SPECjEnterprise2010 – see following charts for details
Example of license + support costs over 5 years:

117
Oracle WebLogic Server costs higher than WebSphere
IBM WAS ND Oracle WLS EE
Runtime performance Best in the industry 5-50% slower than WAS
Flexible mgmt for large deployments Robust framework No -> Extra admin cost
Manage mixed versions in a single cell Robust admin tooling No -> Extra admin cost
Manage DataPower & HTTPD in admin GUI Productive admin tooling No -> Extra admin cost
SOAP & page fragment cache w/ replication Faster performance No -> Extra admin cost
Eclipse toolkit for the Jython admin scripts Productive admin tooling No -> Extra admin cost
Private cloud capabilities Yes via HVE and IWD Limited -> Extra admin cost
SLA, Health and dynamic clustering Included (formerly WVE) No
Production HTTP server included Included No -> Purchase Oracle Web Tier
DB for session persistence included Included No -> Purchase Oracle DBMS
Production LDAP included Included No -> Purchase Oracle Directory
Edge components included Included No -> Purchase Oracle Web Tier
Portlet API (JSR 286), WSRP 2.0 Included for free in WAS No -> Purchase Oracle WebCenter
Native z/OS, Linux on Power Supported No -> Not supported
Communication enabled applications Supported No -> Build your own
Software hypervisor sub-capacity pricing Pay for CPUs used Pay for all CPUs in a server
Warm backup, cold backup (>10 days) Free Purchase full license
Cold disaster recovery site Free Purchase full license
Support cost 1st year is free, 20% thereafter 22% per year, first year is not free

118
Oracle’s practices are not getting good reviews from clients
Source; ITIC Reliability Survey, February 2011: http://itic-corp.com/blog/2011/02/itic-reliabiity-survey-oracle-users-anxiousangry-over-service-support-slippage
Product Performance, Service, Support
(Responded Excellent or Very Good)
78% give high marks to IBM
80
60
40
20
0
Percentage
20% dissatisfied with Oracle
20
15
10
5
0
Product Performance, Service, Support
(Responded Dissatisfied)
Percentage
Last year, over 400 Oracle WebLogic clients, chose IBM WebSphere

WAS vs JBoss, WebLogic, Tomcat (year 2015)

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to WAS vs JBoss, WebLogic, Tomcat (year 2015) (20)

Recently uploaded (20)

WAS vs JBoss, WebLogic, Tomcat (year 2015)