System Sizing

Spreadsheet

EMC | Documentum 6.5

(covers: Webtop, Content Server, Task Space and High Volume Server)

EMC Proprietary and Confidential

Copyright © 1994-2010. EMC Corporation. All Rights Reserved.
Documentum and the Corporate Logo are trademarks or registered trademarks of EMC Corporation in the United S
and throughout the world. All other company and product names are used for identification purposes only and may
trademarks of their respective owners.

Last Update:3/24/2010
WARNING: Information in this tool is likely out-of-date. Please obtain a new version from Documentum.

This Sizing Spreadsheet is for Documentum Version 6.5 only.

180
For assitance with the sizing tool, please email [email protected]
of EMC Corporation in the United States
ntification purposes only and may be

on from Documentum.

Version 6.5 only.

Copyright © 1994-2010. EMC Corporation. All Rights Reserved.

Documentum and the Corporate Logo are trademarks or registered trademarks of EMC
Corporation in the United States and throughout the world. All other company and product names
are used for identification purposes only and may be trademarks of their respective owners.

EMC 6801 Koll Center Parkway Pleasanton, CA 94566 925-600-6800

All other company and product names are used for identification purposes only
and may be trademarks of their respective owners.

The information in this document is subject to change without notice and for internal use only. No
part of this document may be reproduced, stored, or transmitted in any form or by any means,
electronic or mechanical, for any purpose, without the express written permission of EMC
Corporation. EMC Corporation assumes no liability for any damages incurred, directly or
indirectly, from any errors, omissions, or discrepancies in the information contained in this
document.

All information in this document is provided “AS IS”, NO WARRANTIES, WHETHER EXPRESS
OR IMPLIED, INCLUDING THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
FITNESS FOR A PARTICULAR PURPOSE, ARE MADE REGARDING THE INFORMATION
CONTAINED IN THIS DOCUMENT.
This spreadsheet automatically provides estimates of a customer's hardware resource
needs based on user and hardware profiles provided by the customer. Use this spreadsheet as an aid when
you are working with a customer to size a Documentum deployment. The figure below illustrates the spreadsheet's

start

Budget for hardware and

Enter Workload information on software purchases
"Customer Input Page"

Check with hardware budget

and then adjust workload &
Enter Document information on document information
"Customer Input Page" accordingly

Talk with hardware vendor and Inspect Estimated hardware

enter hardware profile on usage from "Output Summary
"Customer Input Page" Page"

The spreadsheet focuses primarily on sizing deployments of the standard Documentum Editions.
It is less useful for deployments that are highly customized or are not primarily made up of
the Edition software.

The hardware resource estimates given in the spreadsheet are derived from Documentum performance benchmar
and analysis. Most of the information that underpins this sheet is summarized in the Documentum
System Sizing guide. Additional detail can be found in the detailed benchmark reports.

Please find updated Sizing Tools and other Sizing materials at the PowerLink and Developer Websites.

EMC Powerlink
Search under : Home > Support > Technical Documentation and Advisories > Software ~ D ~ Documentation > D
eet as an aid when
strates the spreadsheet's use.

m performance benchmarks

Developer Websites.

e ~ D ~ Documentation > Documentum System > Systems Sizing

What's New:
1. Updated D6.5 Numbers
2. Corrected errors in the sheet

Limitations, known bugs, and other notes [updated]

1. Spreadsheet output should be considered suspect at either an extremely low number of users or a very large number of us
Please review large user scenarios with your account team (and the performance team as needed).
2. If you need assistance with the Sizing tool you may contact your Account Manager, Professional Services, or you may sen
3. CPU, Memory and I/O estimates are based on our own workload testing. Actual customer workloads (and usage) could dif
4. Geographic location and separate docbases could imply additional needed servers.
5. This tool assumes that most other machine improvements go hand-in-hand with Mhz improvements. 100% more users can
on an 800 Mhz system vs. a 400 Mhz system.
6. Original benchmarks where performed on 1.2 GHz Sun UltraSparc III processor.
7. This tool assumes that UNIX, including Linux, and NT machines can serve the same number of users with equal processor
8. This tool does not take into account the usage of Vmware.
9. This tool does not take into account temp areas, program installation areas, and operating system areas for disk space cal
10. This tool does not take into account varying bandwidths for client PC to server interactions.
11. This tool does not take into account varying PC machine Mhz. The focus is on server hardware not client hardware.
12. The workloads in this tool did not include any large batch operation that could impact resource consumption. See
detailed benchmark reports for more information on the actual workloads.
13. The hardware resource estimate does not include network bandwidth estimates or estimates for hardware load balancers.
14. The database memory requirements do not accurately account for concurrent user memory needs.
If in doubt add more memory for the database. Also, some recommendations may exceed what a
database could support in a single instance. In those cases it will be necessary to decrease the memory for the RDBMS
by reducing the buffer cache.
15. Spreadsheet does not cover any Documentum Distributed feature (Replication, Content servers, Branch Office Caching S
16. This release does not factor in throughput or space needs of various hardware RAID options.
17. The deployment section of the Output Summary now rounds the number of cpu's per machine up to an even number if gre
18. The deployment section of the Output Summary no longer reports number of cpu's for the Index Server/Agent if the numbe
19. Removed Beta notice.
20. Fixed BPS Calculation displays. Input cells: C70 and Output cells C19, B19, B45, B77
21. Added missing MTS cpu output.
22. Added SQL Server 2005 as a database choice.
23. Updated TaskSpace calculations.
24. Added multi-core support
 a very large number of users.

Services, or you may send an email to [email protected].

ads (and usage) could differ from this.

nts. 100% more users can be served

sers with equal processors and Mhz.

m areas for disk space calculations.

not client hardware.

onsumption. See

hardware load balancers.

emory for the RDBMS

Branch Office Caching Services, etc.).

p to an even number if greater than 1.

Server/Agent if the number of machines is 0.
EMC | Documentum Sizing Input Page WARNING: Information in this tool is likely out-of-date.

Step #1: Enter User/Workload Profile Please obtain a new version from Documentum.

User Profile Webtop 6.5 TaskSpace 6.5

Heavy Users 200 N/A
Light Users 500 0
%Heavy Users Active 100% 0%
%Light Users Active 10% 10%
Heavy Users/Busy hour 200 0
Light users/Busy hour 50 0
Total Users/Busy hour 250 0
Estimated % Growth of Users Per Year 50% 0%
Level of Customization Light None
Workflow Intensive No Yes

Note: While the WebTop and TaskSpace calculations are fine for estimating total server hardware, they should
not be used to compare the 'cost' of using Webtop vs. TaskSpace. The workloads for each are
significantly different. This difference has been accounted for in the hardware sizing,
but does not represent a fair comparison between Webtop and TaskSpace.

Workload-Specific Criteria WDK/Webtop based Applications

Content Server Session Pooling Enabled
Number of custom types 50 Extended HTTP Timeout
Number of CS Instances per machine 1 Clustered App Server
Peak Fulltext Queries per min

no

Step #2: Enter Document Profile Content Loading CPU Input

Content Profile Loading days per year
Num of Original Source Documents: Yr 1 9,000,000 Num. of Docs/Day
Estimated Average Size (kbytes) 1000 Content Input Window (hrs)
Avg. Versions per Document 1 Num. AutoWF Tasks per Doc
Average Additional Renditions 0 Average Size (Kb)
Document or Media
Transformation Services?
Custom Attribute size per Doc (kbytes) 1 Renditioning Priority
Number of Custom Attributes 15 Full Text Indexing

Regular Objects

Document Sizes(kb):

Number of Source
Format/Input Type Average Size (KB) Docs in First Year % of All
Word 1,000 5,400,000 60%
PPT 0 0 0%
PDF 1,000 3,600,000 40%
HTML/Web Pages 0 0 0%
XML/text 0 0 0%
Images 0 0 0%
Contentless 0 0 0%
MPEG 0 0 0%
Total 2,000 9,000,000 100%
Weighted Average 1,000

LightWeight Objects (Requires High Volume Server)

 Document Sizes(kb): Parent

Number of New
Number of Source Docs in Docs Avgerage Object
Input Type First Year Per Year Size (KB)
Example 50,000 1000 25

Step #3: Enter Platform Profile Information

Years of Coverage for Hardware 5 Note: This area is useful in working "WHAT IF?" scenarios
High Availability Needs disaster recovery
Database Server Type Oracle
JVM version 1.6

CPU type MHz Physical CPUs per

server
Web-tier machines Intel_IA64 3000 2
Content Server machines Intel_IA64 3000 2
Index Agent machines Intel_IA32 2400 2
Index Server machines Intel_IA32 2400 2
RDBMS machines Intel_IA64 3000 N/A
BPS Server machines Intel_IA32 2400 2
Site Caching Services Target machines Intel_IA32 2400 2
Document Transformation machines Intel_IA32 2400 2
PDF Aqua Server machines Intel_IA32 2400 2
Media Transformation Servers Intel_IA32 2400 2
EMC Proprietary and Confidential
tion in this tool is likely out-of-date.

w version from Documentum.

Webtop based Applications BPM

on Pooling Enabled Yes Peak Manual Activities per min 0
nded HTTP Timeout No Automatic Activities per hour 0
ered App Server No BPS messages per hour 0
Fulltext Queries per min 10

ent Loading CPU Input

ng days per year 260
30 ** Do not include these documents in the profile below.
ent Input Window (hrs) 24
AutoWF Tasks per Doc 0
age Size (Kb) 2000
ment or Media
sformation Services? None
itioning Priority Low Priority
Text Indexing Immediate ** This fulltext flag is for all workload profile calculations.

Avg Fixml
Number of size KB (if
Request Media New Docs Avg. # of Avg. Rend. Size Average # of Content can be known)
Transformation? Per Year Add'l Rend. (% of Orig) Versions FT Indexed
No 800,000 0 30% 1 Yes 0
No - 0 50% 1 Yes 0
No 1,200,000 0 0% 1 Yes 0
No - 0 40% 1 Yes 0
No - 0 0% 1 Yes 0
No - 0 20% 1 No 0
No - 0 0% 1 No 0
No - 0 15% 1 No 0
0 2,000,000
0 15% 1.0
 Child

Avgerage Avgerage
Number of Content Size Object Size
Children / Parent (KB) (KB) % Materalized
1,000 100 25 20% <--- This row will NOT be included

This area is useful in working "WHAT IF?" scenarios with your hardware vendor

Cores per CPU Planned # of servers disk I/O

non-HA HA capacity
2 1 2 Note: The "planned number of servers" will be
2 1 2 used only to detect if the planned number of servers
2 1 2 you intend to purchase will not meet the capacity
2 1 2 350 demand.
2 1 2
2 1 2
2 1 2
2 1 2
2 1 2
2 1 2
Proprietary and Confidential
EMC | Documentum System Sizing Output Page
User Profile Summary
User population after 5 years 3,544
Users/busy hour after 5 years 1,266
Number of Documents from all sources after 5 years 17,039,000 source
17,039,000 source + versions + rend.
WARNING: Information in this tool is likely out-of-date. Obtain an updated version

Estimated Hardware Resource Summary

Disk
Output CPUs Cores 2 Memory (MB) Space (MB)
Content Server 2 2 13,824 16,677,734
Index Agent/Server 4 6 [ alt. 3.7 5,272 14,367,116
WDK/App Server (Web) 2 2 7,168
RDBMS Server 2 2 2,048 50,768
10
Total for Servers 10 12 28,312 31,095,618

Document Transformation Svr 0 0 Note: These estimates are NOT adjusted for High Availability
BPS Server 0 0
Media Transformation Svr 0 0

Hardware Deployment Options (note: Not Adjusted for HA needs)

Option #1 # of machines CPUs/machine Cores/CPU
Host-based (Web + Content Serv. + FT + DB) 1 10 2
Option #2
Web Tier Server separate 1 2 2
Content Server/FT Index subsystem combined 1 6 2
RDBMS separate 1 2 2

Option #3
Web Tier separate 1 2 2
Content Server separate 1 2 2
Index Agent 1 2 2
Index Server (Full Text Index) 1 2 2
RDBMS separate 1 2 2

Other Servers
Document Transformation Service PCs 0 2 2
BPS Servers 0 2 2
Media Transformation Svr 0 2 2

Full Text Notes

Note: the calculated full text Disk I/O load (402 I/O's per sec) exceeds the entered capacity (350 I/O's per sec)
WARNING: The total num of docs or size exceeds what can be handled by a single Index Server Search Node.
Note: The large full text partition merge could take as long as 9312 min, this may impact save-to-search latency
(c) 2007 EMC Inc EMC Proprietary and Confidential

Adjustments for High Availability

Desired High Availability Option disaster recovery

Option #1 # of machines CPUs/machine Cores/machine

Host-based (Web + Content Serv. + RDBMS) 2 10 2

Option #2
Web Tier Server separate 2 2 2
Content Server/FT Index Subsystem combined 2 6 2
RDBMS separate 2 2 2

Option #3
Web Tier separate 2 2 2
Content Server separate 2 2 2
Index Agent / Index Server (Fulltext) 2 2 2
Index Server (Full Text Index) 2 2 2
RDBMS separate 2 2 2

Other Servers
Document Transformation Service PCs 0 2 2
BPS Servers 0 2 2
Media Transformation Svr 0 2 2

(c) 2007 EMC Inc EMC Proprietary and Confidential

Important Notes
1. These are estimates only. Actual system usage could vary. Please review README sheet.
2. The disk estimates (space and spindles) do not take into account any RAID overhead.
3. The disk estimates (space and spindles) do not take into account work areas, install space, or OS files
4. The memory estimates do not account for OS needs or needs by other applications
Note: Fulltext HA configurations to be supported starting in 5.3 SP1

Please contact PMO for large volume of users

EMC Proprietary and Confidential Rev:40261
versions + rend.
version

Est. Disk
IOs/sec
10
403
133
143

690

r High Availability

I/O's per sec)

r Search Node.
-search latency

See note below

See note below

r OS files
Bulk Load Calculation Page
Warning: DON'T ALTER ANY VALUES ON THIS PAGE

52 loaders -
measured increase due
CPU/Min to to steady # of auto
store state workflow
Num. of Docs/Day 30 103982 docs deletion cpu secs per auto-wf task tasks/doc # of docs CPU min/op
Content Input Window (hrs) 24 Oracle 39 48.750 0.2 0 103982 0.00046883
docs/day (alt 7,692 CS 60 75.000 0.08 0 103982 0.00072128
dmbasic 11.13 13.913 10000 0.00139125

Oracle eCS Loader

CPU minutes needed for batch 0.01406493 0.02163836 0.0417375
batch window CPU mins avail 1440 1440 1440
number of CPUs required 1 1 1

% increase due to steady state

eCS RDBMS deletion
batch window (seconds) 86400 86400 25%
docs per second 0.00034722 0.00034722
disk I/Os per doc 3.26086957 2.53623188
required disk I/O per sec 0.00113225 0.00088064
disk I/Os per sec per spindle 40 40
disk I/Os per sec 0.00113225 0.00088064

Chase's old formula for database disk io = (50*60*21)/10000

Chase's old formula for content server disk io = (337*60*21)/10000