SOE_WAN_Report-2015

STATE OF THE
ENTERPRISE WAN
| 2015

www.aryaka.com Aryaka Networks, Inc. Copyright 2015. Do not reproduce without prior permission.
S t a t e o f t h e E n t e r p r i s e WA N 2 0 1 5 | 1
Executive Summary
The State of the Enterprise WAN is mixed. While certain components of the enterprise WAN are
improving, others are actually hindering the enterprise’s ability to conduct business over the Internet,
especially when application traffic must travel over long distances.
To gauge the State of the Enterprise WAN, we gathered anonymous, aggregated metrics from our own
customer base. Aryaka pioneered the WAN as-a-Service space and delivers network optimization and
application acceleration for the global enterprise to enable better collaboration, communication, and
business productivity.
These services have attracted customers from all corners of the globe. Aryaka’s WAN as-a-Service
solutions connect over 3000 customer sites across 50 countries in all six habitable continents. With
such a large global network, connecting enterprise data centers to branch offices, enterprise sites to
cloud providers, remote mobile workers to headquarters, and enterprise users to web applications in
distant data centers, Aryaka is the only company in the world with the ability to aggregate incredibly
detailed data about what’s happening, end-to-end, over the enterprise WAN.
These factors uniquely position Aryaka to see the end-to-end performance of the Enterprise WAN.
A positive trend we observed is that first- and last-mile broadband links are improving. Enterprises
are also continuing to migrate workloads to the cloud and shift away from on-premises applications
in favor of cloud-based ones, and as they do, they’re learning about how important the WAN is to
application performance.
Other trends are working against the enterprise as information travels over the WAN. Cloud-based
video, teleconferencing, and other rich media gobble up an ever-increasing amount of bandwidth.
Meanwhile, people are clogging the Internet nearly 24 hours a day, since we can now stream movies,
post status updates, check email, and even create graphics-rich presentations from smartphones and
tablets. Add in emerging trends, such as the Internet of Things (IoT) and Machine-to-Machine (M2M)
communications, and, as you can guess, B2B traffic tends to be what suffers the most.
For our second annual report, we investigate several factors that influence enterprise WAN
performance. We look at everything from last-mile investments in APAC to packet loss statistics, from
the most commonly used applications to cloud usage trends and more.

One Network for the Enterprise
WAN as-a-Service
CDN as-a-Service
WAN Optimization as-a-Service
Aryaka’s groundbreaking WAN Optimization as-a-Service solution combines
multi-tenant, purpose-built WAN Optimization technology with enterprise-grade
connectivity via a dedicated, reliable, global core network. Our private core network
is based on globally distributed POPs and innovative first-mile access technology,
along with centralized WAN and application-layer visibility. It delivers LAN-like
performance over the enterprise WAN without the expense of MPLS or the cost and
hassle of WAN Optimization appliance solutions, all in a single, no CapEx monthly
service.
Web Application Delivery as-a-Service
Aryaka’s Web Application Delivery as-a-Service is the world’s only Dynamic Site
Acceleration solution built on a global private core network. Our solution provides
enterprises with stable, consistent, and lightning-fast application performance that
does not fall victim to the vagaries of the public Internet.
Network as-a-Service
Aryaka’s Network as-a-Service combines enterprise-grade, MPLS-like global private
connectivity with proprietary TCP optimization and QoS. Utilizing a dedicated reliable
core network based on globally distributed POPs and centralized WAN visibility, this
service delivers predictable, high-performing network connections that you can deploy
in minutes.
IP Application Delivery as-a-Service
Aryaka’s IP Application Delivery as-a-Service is a state-of-the-art acceleration
solution, engineered to support enterprise mobility needs. The solution combines
enterprise-grade global connectivity with optimization to offer remote and mobile
users a superior way to access centralized enterprise resources.

Table of Contents
1. Introduction: Enterprise bandwidth demand climbs – again | Page 4
2. HTTP traffic continues to dominate – pointing towards heavy cloud adoption rates | Page 8
3. Data reduction improves the performance of key apps | Page 10
4. Cloud-scale technologies are accelerating globalization | Page 12
5. The public Internet is far from business-grade, especially over long distances | Page 17
6. Network problems in the middle mile kill application performance | Page 21
7. Cloud adoption is still on the rise worldwide | Page 23
8. Conclusion: Current trends highlight the need for quick-to-deploy, fully managed | Page 26
enterprise WAN services

Introduction:
Enterprise bandwidth
demand climbs – again
The State of the Enterprise WAN is improving, but
this improvement comes with a series of caveats.
First- and last-mile links are improving globally, espe-
cially as governments in regions such as APAC invest
in infrastructure, but that’s only a small part of the
Internet overall.
The middle-mile, in contrast, is under siege. Demand
for bandwidth is at an all-time high, and we’ll prob-
ably be saying the same thing next year, and the year
after, and so on. Demand is climbing and shows no
signs of slowing down.
Telecommunications giants tend to funnel their
investments towards the first- and last-mile because
they own those pieces of the Internet. The middle of
the Internet, however, is a shared medium, and as
such, investments in it are modest. As bandwidth de-
mands continue to rise, this problem will only worsen.
Today, there are approximately 3 billion people on-
line, or roughly 40 percent of the global population.
That’s up from less than a billion people online in
2005, or less than 16 percent of the world’s popula-
tion.
By 2020, however, we should hit a global tipping
point where more than 66 percent of the world’s
population comes online, or roughly 5 billion people.
This is the claim Peter Diamondis, founder and chair-
man of the X PRIZE Foundation, makes in his new
book Bold.
We agree with his assessment, although we would
argue that with the accelerating adoption of smart-
phones and tablets, Diamondis’ forecast is probably a
conservative one. His estimate also doesn’t factor in
all of the bandwidth being gobbled up by machines, as
the Machine-to-Machine communications (M2M) and
Internet of Things (IoT) trends pick up steam.

Moreover, as ever more business is conducted online,
B2B bandwidth demands continue to climb, as
well. This is common sense. However, the 2000-era
approach to this problem, simply adding capacity,
is still the norm, despite the fact that this legacy
approach is proving to be counter-productive.
What this means for the enterprise is that while
everyone is clamoring for more bandwidth, there’s a
negative and coincident trend we must mention: most
IT budgets are flat or declining.
To make matters worse, the circa 2000-era approach
to this problem, simply adding capacity, is still the
norm, despite the fact that this legacy approach is
proving to be counter-productive.
In recent years, researchers have studied the effect
of adding more highway capacity in cities in order to
ease congestion. The result? Congestion remains the
same. More roads actually encourage more driving.
The same is true of our Internet Superhighways.
Where the comparison to highways stops, though,
is that the enterprise has the ability to solve the
problem. City commuters must rely on government
entities to build out and maintain transportation
infrastructure. The enterprise does not need to.
What the enterprise must cope with is monopoly
telcos that prioritize profit margins and their
monopoly fiefdoms over all else.
As we compiled data for our report, we saw network
traffic growing across all regions. For instance, North
America bandwidth growth, while still high, is now
outpaced by growth elsewhere in the world – as the
rest of the world comes online at an accelerated rate.
To bypass those crowded, under-maintained public
Internet highways – which kill the performance of
cloud applications – most large enterprises rely on
their own private highways, which in the past typically
consisted of a combination of Multiprotocol Label
Switching (MPLS) plus expensive WAN Optimization
hardware.
Here’s the problem with that approach: it runs
counter to another key enterprise trend, the
widespread adoption of cloud services.
Figure 1 : Traffic Growth by Region: warning signs of a looming cloud crunch?
102%
173%
201%
399%
190%
236%
67%
64%
69%
110%
38%
63%
0% 50% 100% 150% 200% 250% 300% 350% 400% 450%
India
China
APAC
EMEA
North America
Overall
Median Mean

The cloud is supposed to bring down costs, deliver
agility, and enable a more nimble, flexible enterprise.
If you have to travel over super expensive private
networks and deploy equally expensive and
complicated optimization hardware, the cost needle
starts to point in the wrong direction.
MPLS unit prices are not likely to decline year-over-
year to stay in line with flat budgets and growing
demands for capacity. In fact, few enterprises are
lucky to see MPLS costs declining by 10-15 percent
per year, if at all. If this trend continues – and there
is no reason to think it will not – smarter alternatives
to private MPLS are needed to keep up with traffic
growth across the new distributed, global, cloud-
driven enterprise.
Simply put, doubling the bandwidth of expensive
private WAN services like MPLS every 12-18 months
is just not an option for most businesses.
And if we move our focus from North America to the
rest of the world, these legacy approaches are non-
starters.
One of the major traffic trends we observed is a rise
in traffic from India, China, and other APAC locations
accessing North American and EMEA data centers
and cloud instances.
Judging from the traffic we observed, enterprises of
all sizes in all sectors of the economy are increasingly
looking for innovative ways, other than MPLS and
VPNs, to connect with overseas sites.
Part of why this is happening is that, contrary to
conventional wisdom, bandwidth at non-headquarters
enterprise sites, which we refer to as “access sites” in
this report, is actually reasonably good worldwide.
Even in developing countries, access-site bandwidth
is almost never below 2 Mbps (fewer than 3 percent of
those links are below that threshold).
The chart below shows the average access WAN link
size at locations around the globe.
(Note: For the WAN link charts below, we eliminated primary headquarter and primary data center sites,
which generally have high-bandwidth data links and would skew the data. The sites represented are the
remote access sites, such as branch offices, that are accessing applications located at headquarters and other
centralized data center locations.)
Figure 2 : Average global “access site” WAN link speeds
(For a breakdown on each region, refer to section 4.)
3%
19%
12%
26%
17%
8%
9%
7%
0%
5%
10%
15%
20%
25%
30%
<2 2-4 5-9 10 11-20 21-40 41-99 100
Global

If we drill down further to determine the variation
in demand for bandwidth across industries and
verticals, we see a rather uneven distribution. While
some sectors saw a steep surge in bandwidth demand
compared to last year, others grew more slowly.
The software vertical saw the highest growth in
bandwidth demand, with demand more than
doubling compared to last year. Manufacturing and
Computer Hardware companies too saw a huge surge
in bandwidth demand. Demand for bandwidth grew
in the financial sector as well. Bandwidth demand in
other sectors including Technology, Healthcare, and
Logistics remained more or less similar to that of last
year.
Figure 3 : Average bandwidth demand across Industries
296%
307%
127%
111%
177%
62%
142%
67%
32%
3%
0% 50% 100% 150% 200% 250% 300% 350%
Manufacturing
Software
Computer Hardware
Financial
Technology

88%
80%
79%
68%
51%
37%
35%
33%
27%
26%
17%
16%
13%
11%
HTTP
HTTPS
CIFS
MS Remote Desktop
EPMAP
SMTP
SSH
FTP
DNS
MSSQL
Citrix
LDAP
Print JetDirect
MS DFS Replication
HTTP and HTTPS traffic
continues to dominate – pointing
to heavy cloud adoption rates
Keeping pace with what we saw in last year’s report,
HTTP and HTTPS traffic continues to dominate. What
this indicates is, that applications are continuing to
transition away from on-premises data centers, to
web-based consumption and delivery models.
Developers and researchers have been forecasting this
shift, and our data backs those predictions up.
The chart below shows the percentage of companies
that use various applications. In other words, 88
percent of our enterprise customers send HTTP traffic
over the WAN; 80 percent of them send HTTPS, while
only 11 percent send specific application traffic like
Microsoft DFS Replication.
(Note: For the application charts below, we eliminated applications that constituted less than 10% of traffic
across the network)
Figure 4 : Application Usage

The next chart shows the percentage of traffic that
each protocol or application represents overall. In
other words, for those 88 percent of companies
sending HTTP traffic, an average of 23 percent of all
traffic that each company individually sends over the
Aryaka WAN is HTTP traffic. Adding HTTPS to this
metric, many companies are seeing that 40 percent or
more of their WAN traffic is HTTP-related.
For companies running CIFS (Common Internet File
System, which is typically used for file access and
remote printing in Microsoft Windows environments),
25 percent of their overall WAN traffic is CIFS. In
other words, for those 79 percent of companies
sending CIFS traffic, a quarter of all traffic that each
company individually sends over the Aryaka WAN is
CIFS traffic.
Figure 5 : Top Applications as a Percentage of Traffic
24.8%
22.6%
21.6%
16.0%
12.6%
10.4%
4.7%
4.0% 3.8%
0.4%
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
CIFS HTTP HTTPS FTP MS RDP SSH SMTP MSSQL EPMAP DNS

S t a t e o f t h e E n t e r p r i s e WA N 2 0 1 5 | 1 0
Data reduction improves the
performance of key apps
As Big Data applications continue to grow in
popularity, it’s no surprise that the demand for data
is at all-time high, and could well increase at an
exponential rate in coming years. The primary drivers
for this increased demand for all types of data include
globalization, increased cloud adoption, mobility,
richer apps, and streaming videos, to name a few. In
fact, many of these apps now have some sort of data
collection and analytics features built into them.
Therefore, the only feasible solution enterprises are
left with to counter this data explosion is figuring
out how to navigate through middle-mile congestion
efficiently, and the primary tool for this is WAN
Optimization.
The principal goal of any WAN Optimization solution
is to maximize the efficiency of the data flow across
the enterprise network. Different vendors have
different approaches to achieving that goal, but data
reduction technologies and techniques remain at the
core of any solution.
Among the many techniques leveraged in WAN
Optimization solutions, data deduplication,
compression, and caching account for a sizeable
reduction in data flowing across a WAN pipe. Using
these technologies, WAN Optimization vendors are
able to provide not just bandwidth savings, but also
the ability to accelerate application performance.
The data shows that, despite the talk around more
application data being encrypted or pre-compressed,
many protocols, including CIFS, HTTP, IMAP, and
MSSQL, can still see bandwidth usage savings of
50 percent or more. With the right optimization
technologies in place, data reduction can help
enterprises keep up with skyrocketing bandwidth
demands, at least in the short term. In the longer
term, more advanced, unified approaches will be
needed. Another noteworthy point here is that certain
applications are far more compressible than others.
Perforce, for instance, sees an average data reduction
of 75 percent over our network.

81%
75%
71% 70%
66%
60% 60% 58%
51% 50%
48% 48%
44% 44%
41% 40%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Figure 6 : Data Reduction

Cloud-scale technologies are
accelerating globalization
It’s not just big enterprises that are taking advantage
of globalization trends anymore. Businesses of all
sizes are outsourcing everything from call center
support to general customer service to R&D.
At Aryaka, we are in a unique position to watch this
trend evolve, since our flagship technology, WAN
Optimization as-a-Service, was originally developed
to help enterprises connect to distant, and often
overseas, branch offices.
In fact, our entire portfolio of services is built on top
of a private network made up of a global footprint of
POPs, which are within 30 milliseconds of 90 percent
of the world’s enterprise users. Aryaka’s WAN as-a-
Service solutions connect over 3000 customer sites
across 50 countries in all six habitable continents.
What we noticed long ago is that even large
enterprises want to move away from outdated
technologies like MPLS, WAN Optimization
appliances, and commodity CDNs. For them, the
reasons are usually more about flexibility and agility
than simply cost. For smaller businesses investigating
these services for the first time, they now have access
to technologies they simply couldn’t afford before.
Quite a few businesses are moving beyond IP VPNs
to an enterprise-grade private network that has
optimization built in to it, so as to mitigate the effects
of latency, congestion, packet loss, and jitter. In other
words, application performance for all users – no
matter where in the world they are located – is the
primary driver.
The use case for our initial customers was typically
to use our network to tackle globalization challenges,
accelerating their application traffic to employees,
customers, and partners located far away from
centralized servers. Fast forward to today, enterprises
also use Aryaka’s technology to navigate far more
complex use cases such as enterprise mobility and
accelerating access to their workloads in public,
private, and hybrid clouds.

In 2014, we again observed regional Internet
bandwidth and speeds improving globally. Data shows
that 67 percent of our customers use a high-speed
Internet link (greater than or equal to 10 Mbps) for
their last mile.
Meanwhile, 16 percent of our customer base uses
very-high-speed Internet edge links (greater than 40
Mbps). On the negative side, congestion and packet
loss on the unreliable Internet results in sub-optimal
use of available bandwidth, so there is still plenty of
room for improvement, even on the best parts of the
Internet.
Broken down by region, EMEA locations have the
highest percentage of high-speed access site last-
mile WAN links. But, in line with last year’s trend
and owing to the improved quality of last-mile links
globally, EMEA now lags behind the global average for
very high-speed WAN links.
(Note: For the WAN link charts below, we eliminated primary headquarter and primary data center sites,
which generally have high-bandwidth data links and would skew the data. These are the access sites, such as
branch offices, that are accessing applications located at headquarters and centralized data center locations.)
Figure 7 : Average worldwide access site WAN link speeds
Figure 8 : WAN Link Size (Mbps) - EMEA
3%
19%
12%
26%
17%
8%
9%
7%
0%
5%
10%
15%
20%
25%
30%
<2 2-4 5-9 10 11-20 21-40 41-99 100
Global
1%
15% 15%
29%
13% 13%
6% 7%
0%
5%
10%
15%
20%
25%
30%
35%
<2 2-4 5-9 10 11-20 21-40 41-99 100
EMEA

North American access sites, while lagging behind
EMEA slightly in regards to high-speed last-mile
links, actually have a higher percentage of very high-
speed links.
Compared to last year, APAC has seen a substantial
growth in very-high-speed WAN links, even
surpassing EMEA and the global average. A full 21
percent of APAC sites have links >20 Mbps, up from
14 percent last year. Another thing to note in APAC
is that many sites (about 33 percent) still have low-
speed connections of less than 10 Mbps. This indicates
a huge amount of disparity among enterprises with
offices in APAC.
Figure 9 : WAN Link Size (Mbps) - North America
Figure 10 : WAN Link Size (Mbps) - APAC
7%
20%
9%
23%
16%
8%
10%
8%
0%
5%
10%
15%
20%
25%
<2 2-4 5-9 10 11-20 21-40 41-99 100
North America
1%
20%
12%
26%
19%
6%
9%
6%
0%
5%
10%
15%
20%
25%
30%
<2 2-4 5-9 10 11-20 21-40 41-99 100
APAC

While China and India are included in our overall
APAC numbers, we also decided to look into each one
of them separately. We believe stand-alone data from
these high-growth, emerging economic behemoths
has significance and points to globalization patterns
that could be missed in regional data. Patterns for
China mirror those of the APAC region as a whole,
showing considerable growth in very high-speed WAN
links and significant disparity.
India also saw an increase in the percentage of very
high-speed WAN links; however, 23 percent of Indian
enterprises still rely on <4 Mbps WAN links.
Figure 11 : WAN Link Size (Mbps) - China
Figure 11 : WAN Link Size (Mbps) - India
1%
20%
12%
26%
19%
6%
9%
6%
0%
5%
10%
15%
20%
25%
30%
<2 2-4 5-9 10 11-20 21-40 41-99 100
China
5%
18%
19%
23%
16%
5%
15%
0%
0%
5%
10%
15%
20%
25%
<2 2-4 5-9 10 11-20 21-40 41-99 100
India

Looking at only high-speed and very high-speed last-
mile access site connections, EMEA has the highest
percentage of sites with connections greater than 10
Mbps. However, it has the lowest percentage of very
high-speed (>40 Mbps) sites.
Worldwide, the percentage of sites in the 11-40 Mbps
range is 24 percent, down from 29 percent last year,
but we believe that most of them have upgraded to
very high-speed links, as we observed an equivalent
increase in very high-speed sites compared to last
year. 16 percent of sites worldwide, up from 12
percent last year, enjoy very high-speed connections
of greater than 40 Mbps.
Thus, globally, high-speed plus very high-speed
connections account for 40 percent of all sites
connecting to the Aryaka network. This is similar
to what we saw last year. Sites with 11-40 Mbps are
fairly consistent worldwide, accounting for roughly a
quarter of all sites, no matter which region you look
at.
The chart above tabulates the WAN link sizes for our
top 5 verticals. Companies in the Technology sector
have the highest demand for very high-speed WAN
links, while Logistics companies are skewed towards
medium-to-low speed links, typically 10 Mbps or
less. More than half of the links in the Healthcare,
Software, and Technology sectors rely on high-speed
or very high-speed WAN links. Overall, Software
companies have the highest demand for high-speed
WAN links.
Table 1 : EMEA has the highest percentage
of high-speed access site links
Table 3 : EMEA has the highest percentage of high-speed access site links
Table 2 :North America has the highest
percentage of very high-speed access site links
Region/Country
%Above 10 Mbps
(<40Mbps)
Overall 24%
EMEA 26%
APAC 25%
North America 24%
China 23%
India 21%
Region/Country %Above 40 Mbps
Overall 16%
China 18%
North America 18%
APAC 16%
India 15%
EMEA 13%
Healthcare Logistics Manufacturing Software Technology
<2 0% 10% 2% 0% 0
2-4 5% 46% 17% 7% 4%
5-9 5% 10% 11% 14% 13%
10 33% 20% 32% 20% 23%
11-20 24% 3% 16% 36% 19%
21-40 19% 3% 9% 11% 4%
41-99 5% 2% 9% 7% 19%
100 10% 5% 5% 5% 17%

78%
14%
4% 4%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
Global
The public Internet is far from
business-grade, especially over
long distances
For this report, Aryaka measured anonymized
enterprise WAN traffic from 3,000 customer sites
using the Internet links between enterprise locations
and the nearest Aryaka POP in order to measure the
quality of last-mile Internet connections. Aryaka has
partnered with Tier 1 ISPs, so that our POPs have
multiple, redundant high-bandwidth Internet pipes
that branch offices can connect to. Therefore, network
performance between Aryaka POPs and end user
locations, most of which are within 30 milliseconds or
less from the nearest POP, is an excellent indicator of
last-mile Internet link quality.
Aryaka’s architecture generates ping-based
measurements on each link once per second, which
is an efficient and optimal method to measure packet
loss, especially for small packets. It also measures
TCP retransmissions, which is a good way to measure
packet loss for larger packets. Here’s what we found:
Figure 12 : Last Mile Link Quality

Last-mile link quality is measured in units of “bad
minutes.” We define a “bad minute” as one where,
either measured ping packet loss exceeds 4 percent
or where ping packet loss exceeds 1 percent and
TCP retransmissions exceed 3 percent. During a
bad minute enterprises will not get the application
performance that they need to function smoothly at
optimal productivity levels.
Experiencing a bad minute every once in a while –
say, less than 0.5 percent of the time – is unlikely to
cause issues in delivering high-quality, consistent
application performance. The more frequently bad
minutes occur, however, the greater the risk that
users will frequently report application performance
problems.
Globally, 78 percent of the sites overall experienced
<0.5% bad minutes.
As one might expect, the percentage of sites
experiencing a relatively large amount of bad minutes
in regions like North America and EMEA is low. A big
surprise when comparing data to last year, however,
is the fact that last-mile link quality has actually
degraded for EMEA.
84%
12%
3% 2%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
North America
86%
6%
2% 5%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
EMEA
Figure 13 : Last Mile Link Quality - North America
Figure 14 : Last Mile Link Quality - EMEA

We have included India and China in our APAC
numbers, but we also decided to look into each one of
them separately.
APAC has seen a sound improvement in the quality
of last-mile links, with bad minutes of >2 percent
dropping significantly from a quarter last year to just
about 12 percent this year.
The improvement in quality of last-mile links in China
is considerable, which pulls up the APAC numbers
overall. Bad minutes greater than 2 percent saw a
steep drop from 32 percent last year to just 10 percent
this year.
While the last-mile link quality improved the world
over, the same cannot be said for India. Even though
the percentage of bad minutes greater than 2 percent
remained about the same as last year, the chart
this year is skewed towards unacceptably high bad
minutes of greater than 5 percent.
70%
18%
5% 7%
0%
10%
20%
30%
40%
50%
60%
70%
80%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
APAC
67%
23%
5% 5%
0%
10%
20%
30%
40%
50%
60%
70%
80%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
China
Figure 15 : Last Mile Link Quality - APAC
Figure 16 : Last Mile Link Quality - China

The above charts indicate that, overall, unaided last-
mile business Internet connectivity is reasonably good
in most places around the globe. This suggests that
the last-mile edge connection is not an inhibitor to file
transfers, email usage, surfing the web, and accessing
nearby applications.
Also, if you look at the charts above, enterprise sites
in the first column are seeing what is essentially L2 or
MPLS-like last-mile quality, which confirms our thesis
that last-mile links across short distances behave very
similar to private links with dedicated bandwidth in
most parts of the world. Application performance over
the Internet when traffic crosses oceans is entirely a
different matter, however, since the public Internet
tends to perform poorly over long distances owing to
high levels of congestion and packet loss.
For APAC overall, the number of unacceptable bad
minutes experienced at various locations is still high,
with 7 percent of sites experiencing bad minutes
more than 5 percent of the time. The number of
sites experiencing high numbers in terms of bad
minutes (2-5 percent) has gone down dramatically
from 16 percent last year to a mere 5 percent this
year, showcasing the significant investments APAC
governments are channeling in to telecommunications
infrastructure.
India is not keeping pace with the rest of the region,
however, with a combined 20 percent of all sites
having poor last-mile connections that need some
form of optimization – even over the short last-mile
hop.
To conclude, overall unaided last-mile Internet quality
without optimization is decent for almost all global
locations. For some locations however, especially in
APAC, last-mile connectivity that is not optimized is
still a source of application performance issues.
60%
20%
4%
16%
0%
10%
20%
30%
40%
50%
60%
70%
< 0.5% Bad minutes 0.5% - 2.0% Bad
minutes
2.0% - 5.0% Bad
minutes
> 5.0% Bad minutes
India
Figure 17 : Last Mile Link Quality - India

Network problems in the middle
mile kill application performance
Latency and bandwidth limitations are not the only
bad guys when it comes to delivering acceptable
application performance over the Internet. Congestion
and packet loss are the real villains, and the public
Internet is their breeding ground. Congestion and
packet loss aggravate the effects of latency when it
comes to moving data across long-haul international
links. In fact, congestion-based packet loss problems
affect application performance even more, as the
distance between the client and server grows. Even
though Internet quality continues to improve on the
last mile, the middle mile is, more often than not,
the true bottleneck. The problem can be traced to
congested ports or peering points where local ISPs
hand off traffic to other networks. In other cases,
such as MPLS, the speed of light barrier still limits
throughput and performance over the WAN.
Last-mile issues combined with the equivalent of a
cobblestone-paved middle mile result in application
performance that is not viable for today’s agile
enterprise ecosystem. Productivity takes a hit, costs
go up, and profit margins shrink as a result of poor
application performance.
Furthermore, as more and more applications as
well as enterprise workloads move from behind the
firewall and into the cloud, traditional networks and
optimization technologies fail to deliver on their
promises, causing the situation to worsen. This is
what’s causing the most worry for CIOs, IT Directors,
and Network Managers today.
Let’s dig deeper into the packet-loss statistics. The
numbers in the next chart indicate that most last-mile
Internet connections work well. Enterprises typically
see less than 0.25 percent packet loss on the last-
mile link, even in places like China and India. The
global median packet loss is a mere 0.04 percent. In
practice, TCP applications can handle packet loss
rates of about 1% over any given time period before
application performance seriously degrades. And the
very nature of TCP is designed to cause some packet
loss in order to efficiently use available bandwidth.

However, many Network Managers assume (wrongly)
that if packet loss is less than 1 percent on their
enterprise WAN, things are fine. The reality could be
very different – especially if you ask the application
end users. Unfortunately, IT usually finds this fact out
the hard way.
Once you drill down into specific network metrics,
network quality varies greatly across regions, and it
can vary within those regions as well, depending on
the diversity of providers in the region and how well
they peer with one another. It can also vary depending
on the time of day. For instance, quality takes a
serious hit during business hours, as the amount of
congestion increases.
The link quality data here only represents the last-
mile packet loss. But the last mile usually isn’t the
problem, since the farther packets travel, the more
middle-mile loss (in addition to any last-mile loss)
you can expect, even over private circuits like MPLS.
More importantly, the more packet loss that occurs
anywhere on the network, the more end users will see
application performance drop.
If you deliver applications hosted in North America
to an end user based out of India or vice versa, the
performance will be significantly lower than if the
same application were delivered over the LAN or even
within the continent.
The traditional way to avoid the effects of a congested
WAN starts with simply adding more capacity. For
more complex WAN topologies (and for organizations
with higher budgets), strategies include optimizing
traffic through compression, de-duplication,
application-specific acceleration proxies, and other
WAN Optimization techniques.
However, when done only at customer end points,
these optimization solutions have limits in terms of
how much they can address performance over the
WAN. And when one end point is in a public, private,
or hybrid cloud, an appliance-based solution rarely
works. Adding boxes to the enterprise networking
ecosystem also adds complexity, CapEx costs
related to the hardware, and ongoing support and
maintenance efforts.
A superior approach is one delivered from an
optimized, intelligent, cloud based WAN as-a-Service
platform, such as the one delivered by Aryaka, which
we’ll discuss in our conclusion.
1.01%
0.65%
1.37%
1.15%
1.06%
2.20%
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
Overall North America EMEA APAC China India
Mean Median
Figure 18 : Packet loss per region

Cloud adoption is still on
the rise worldwide
Cloud adoption continues to rise worldwide, now
more rapidly than ever. Last year, we noticed an
interesting trend: a notable segment of our enterprise
customer base was adopting cloud services, and of
those early adopters, many were already investing
in accelerated network access to cloud services to
provide a seamless experience to all their global users.
In 2014, that trend continued. More than 15 percent
of our customers are now accelerating access to cloud
services via Aryaka. If this trend remains steady,
the cloud will become the default way of delivering
applications in future. Along with an increase in raw
adoption, we have also observed a significant increase
in the willingness to adopt private – and hybrid –
cloud models. Enterprises fear that if they do not start
figuring out a workable cloud strategy now, they run
the risk of losing out to their cloud-savvy competition.
Companies using
accelerated cloud
access
15%
Others
85%
Companies using accelerated
cloud access
Others
Figure 19 : Percentage of Companies Using Accelerated Cloud Access

It’s important to note, as you read this report, that the
cloud-access section is not necessarily representative
of the overall enterprise cloud migration and cloud-
usage patterns. In fact, we believe our customer base
foreshadows trends that will spike in coming years.
Remember, Aryaka is a cloud-based WAN as-
a-Service company. Our customers, obviously,
aren’t reluctant to switch to cloud-based tools that
outperform legacy technologies. In fact, considering
our unique value proposition of providing accelerated
network access to any cloud service, our customers
could even be considered aggressive early adopters of
cloud services, since they prioritize cloud performance
and performance predictability in their IT objectives.
While our customers are clearly blazing the path
to cloud adoption and are ahead of the curve, they
provide a glimpse of the future, showing us all where
the enterprise WAN is headed.
We are not alone in the belief that our customers are
foreshadowing tomorrow’s status quo. Analyst firms
and major technology incumbents have identified
complementary trends in their various studies. For
instance, IDC in its Cloud Predictions for 2015 has
revealed that more than 65 percent of enterprise
IT organizations will commit to hybrid cloud
technologies before 2016.
Cloud adoption is on the rise across market sectors
too. Gartner’s latest estimates (Forecast: Public Cloud
Services, Worldwide, 2012-2018, 4Q14 Update) find
that the global public cloud market is expected to
reach at least $191 billion by 2020.
The next chart, then, is a deeper dive into the traffic
patterns of the enterprises that are accelerating cloud
access over Aryaka. The chart shows the percentage of
cloud traffic versus all other WAN traffic.
One interesting observation to note here is that the
companies that have a high percentage of cloud
traffic relative to total WAN traffic are also investing
in acceleration services. From this, it’s fairly easy to
conclude that network performance problems are
holding back cloud app adoption. Nearly 49 percent
of all traffic from these companies is accelerated cloud
traffic, implying that these companies are serious
about their investments in cloud services.
48.8%
51.2%
Accelerated Cloud Traffic
All other traffic
Figure 20 : Accelerated Cloud Traffic - Mean

As we dug deeper into the data, we found that the
median percentage of cloud traffic vs. all other traffic
was actually 50.5 percent, a bit higher than the 48.8
percent average shown above. A probable reason
for this slight difference is that companies that are
currently in the process of transitioning from on-
premises applications to cloud, will only have a very
small percentage of their total traffic in the cloud,
which slightly skews the results.
Another important thing we should note here is
that the cloud traffic in the chart represents only
accelerated cloud access. If a company is using, say,
Salesforce.com or Google Apps and routing that traffic
directly to the Internet from their locations instead
of routing it through the Aryaka core, that traffic
will not have been counted in this percentage. This
implies that an even higher percentage of the total
WAN traffic for these enterprises might be going to
cloud services. The reason why enterprises have such
a significant percentage of accelerated cloud traffic is
because most cloud services are centrally hosted and
hence need optimization in order to provide a smooth
experience to users across geographies.
50.5%
49.5% Accelerated Cloud Traffic
All other traffic
Figure 20 : Accelerated Cloud Traffic - Median

Conclusion: Current trends highlight
the need for quick-to-deploy, fully
managed enterprise WAN services
IT organizations today are under more pressure
than ever. CIOs and IT managers must cope with
shrinking CapEx budgets, the rising adoption of
cloud services, globalization, shrinking IT workforces
but continuously increasing workloads, and the
expectation that IT can handle everything from Big
Data applications to security to network monitoring
– all while ensuring a great end user experience. Add
it up, and IT needs quick, easy-to-deploy, easy-to-
manage (or zero management) solutions.
On the networking side, IT needs to find flexible and
scalable networking solutions that can keep pace with
the demands of a rapidly growing enterprise user base
and that can support any sudden increase in data
volume. The old model of adding more bandwidth,
deploying MPLS, or installing optimization appliances
does not work anymore. It is labor intensive and
requires IT to continuously focus on network
maintenance and management instead of spending
time on innovative and strategic projects.
As more workloads and applications move to the
cloud, the public Internet itself has become a major
bottleneck. Users have come to expect LAN-like
speeds – for everything – and when an organization
fails to deliver, user satisfaction and productivity
drops. Even the biggest enterprises with the deepest
pockets will not be able to pay for point-to-point
private lines to every single workload and application
they use in the cloud, from Microsoft Azure to
Amazon Web Services to Office 365 to Salesforce.com
to industry-specific applications. You would need a
gigantic budget to keep up.
To ensure a smooth transfer of all enterprise
applications and workloads to the cloud, the focus has
now turned to the WAN and how it can be overhauled
to accommodate the dynamic requirements of the
modern enterprise. The unaided Internet last-mile
links may not be reliable and predictable, but they
have improved significantly over the last decade.
However, as discussed above, when traffic crosses
regions or travels to offices an ocean away, application
performance drops significantly.

When middle-mile congestion problems are factored
in, combined with globalization and rising cloud
adoption, enterprise WAN performance suffers even
more. Growing bandwidth demands aren’t being met
with an increase in IT budgets, nor is the cost of MPLS
coming down quickly enough to allow IT budgets to
keep pace.
The growing diversity of cloud workloads and
increasing M&A activities are two more factors
that make it difficult for the enterprise to predict
traffic patterns. IT is also already struggling to cope
with ever-increasing end-user demand for better
performance and faster response times, whether
those end users are employees, customers, partners,
contractors, or anyone else granted access to
centralized web, cloud, or on-premises applications.
With fixed bandwidth MPLS links that have long
deployment times, it becomes nearly impossible for
enterprise CIOs to keep up with the fast-changing
needs of their organizations.
What we found when analyzing the traffic traveling
over the Aryaka network is a confirmation that a
multi-segment architecture enables enterprises to
address the congestion and quality issues many
last-mile Internet links face, while avoiding Internet
middle-mile issues completely and achieving superior
performance over long distances.
Enterprises want an agile WAN architecture that can
be provisioned fast and is cost-effective.
Aryaka delivers just that. With a globally distributed
footprint of POPs situated close to end user
locations, a dedicated private core network with
built-in optimization, and redundant network and IP
connectivity from top tier network providers, Aryaka
answers the challenge through its multi-segment
architecture.
Aryaka’s multi-segment architecture delivers the
following benefits:
• Virtually eliminates middle-mile packet loss
• Minimizes the amount of last-mile loss
• Minimizes the impact of last-mile loss on
application performance, when it does occur
• Works equally well for enterprise site-to-site as
well as site-to-cloud service/SaaS connectivity
These factors, among others, are why Gartner has
positioned Aryaka as a “visionary” in its Magic
Quadrant for WAN Optimization Controllers for three
consecutive years.
Approaches like Aryaka’s WAN as-a-Service
platform are needed to ensure acceptable application
performance on a future-proofed enterprise WAN.

About Aryaka
Aryaka, the pioneer of WAN as-a-Service, delivers network optimization and application
acceleration for the global enterprise, for better collaboration, communication, and business
productivity. Aryaka eliminates the need for costly WAN appliances and long-haul private
links. Aryaka’s groundbreaking WAN Optimization as-a-Service accelerates any application
to any location and provides organizations with an affordable, optimized, and fully meshed
network to access on-premises applications, as well as cloud services. Aryaka’s Application
Delivery as-a-Service enables globally distributed employees, customers, partners, and
mobile end users to quickly access centralized enterprise applications and public-facing web
resources from anywhere in the world. All services provide end-to-end visibility with 24×7
world-class support.
www.aryaka.com

SOE_WAN_Report-2015

More Related Content

What's hot(20)

Similar to SOE_WAN_Report-2015(20)

SOE_WAN_Report-2015