How to Pick the Right WAN Optimization Solution for Your Organization

How to Pick the Right WAN Optimization Solution for Your ...
http://www.gartner.com/technology/reprints.do?id=1-1FCG...
How to Pick the Right WAN Optimization
Solution for Your Organization
3 April 2013 ID:G00250203
Analyst(s): Bjarne Munch
VIEW SUMMARY
STRATEGIC PLANNING ASSUMPTION
WAN optimization controllers continue to have significant differences in functionality. Network managers
who choose the wrong solution for their environments will have to deal with unresolved performance
issues, suboptimal performance resolution and higher costs.
The WAN optimization market will remain in constant
and significant change until at least 2016.
EVIDENCE
1 The evaluation of vendors' support for specific
functionality is based on a survey of the vendors
included in the Gartner "Magic Quadrant for WAN
Optimization Controllers," as well as a detailed review
of these vendors' product descriptions and data sheets.
Overview
Key Challenges
Although the WAN optimization market is mature, with a large number of viable solutions, there
remain differences among all vendors in their support for all functional areas and their adoption of
new functionality, making it a continued challenge for network managers to stay informed.
Mapping the evolving vendor solutions to specific enterprise needs and use cases can be
challenging, given that all vendors claim to provide a panacea for all environments.
Recommendations
Do not choose a WAN optimization solution based on a preconceived vendor preference, as even
leading vendors are not leaders in all functional areas.
Be sure you have current information, including product road maps, about WAN optimization
products, as this is a constantly evolving market.
Define objectives and functional requirements as the first step in selecting a WAN optimization
vendor. Then create your WAN optimization shortlist based on a vendor's ability to meet these
functional requirements.
Perform a live trial of preferred products before committing to the purchase. Participate actively in
the installation and configuration of each proof of concept to understand the complexity of
deploying each solution.
Introduction
WAN optimization is a vibrant market, equal to more than $1.2 billion in annual revenue. The Gartner
Magic Quadrant for WAN Optimization Controllers for 2012 includes eight appliance vendors, but several
other vendors are viable choices. Despite being a large and fairly mature market, there is a significant
variation in the functions each vendor supports.1 For the last eight to 10 years, WAN optimization
controllers (WOCs) have been a product area consuming similar and adjacent functions, such as
application acceleration, caching, compression and deduplication, traffic visibility, control,
measurements, content delivery network (CDN), video delivery and WAN path control. This trend of
expanding product functions is likely to continue until at least 2016.
While this trend offers network managers an opportunity to simplify their solution deployments, it also
provides a challenge to ensure that your company selects the right product with the right functionality.2
In this research, we outline key functional evaluation criteria to ensure that network managers select
the right product (see Table 1).
Table 1. Summary of Select Use Cases and Related Issues and Solutions
Use Case
Key Objective
Reduce WAN cost
To reduce WAN costs by reducing bandwidth usage
Key WOC Functions
Compression and deduplication
Caching
Optimizing streaming video
delivery
Network traffic management
Enable data center
consolidation
1 of 5
To ensure that application performance remains
acceptable after the application has been moved from
a decentralized deployment to a centralized
deployment
Application acceleration
Network traffic management
2 This is evident via recurring client inquiries about
WAN optimization. A typical client inquiry question is:
"Which vendors should I look at if I want to solve this
problem…?" In other inquiries, clients have told us they
chose a WOC based on a preconceived vendor
preference, but then, for example, found that problems
related to a poorly performing SharePoint were not
solved, because the vendor did not support SharePoint
optimization. We have seen clients shortlist two
vendors with a price difference of close to 50%, where
the vendor with a low-priced product was just as good
a fit as the vendor with an expensive offering.
NOTE 1
COMPRESSION VERSUS DEDUPLICATION
Compression and deduplication are similar. Both
methods analyze the data before transmission for
repetitive byte patterns and replace them with small
references for the transmission:
Compression identifies redundancy in a specific file
being transmitted and then removes those
redundancies. Typically, standard algorithms such
as gzip and LZ are used for compression.
Deduplication works by identifying redundancy
within all files being transmitted, and the algorithm
looks for longer byte patterns than compression.
Typically, proprietary methods are used for
deduplication. Because deduplication works
generically across all files identifying long patterns,
significantly higher compression is possible;
however, larger memory is also required.
NOTE 2
VIDEO STREAMING
Depending on the video type and format, the bit rate
for a video stream can range from 375 Kbps up to 6
Mbps for HD video. If several employees are watching
a stream at the same time, the network impact is
equally higher (e.g., 10 people watching a 375 Kbps
stream add more than 3 Mbps). To reduce network
impact, vendors implement media servers to manage
the media stream. This needs to be done for each
streaming protocol in the network. There are several
dedicated streaming protocols — e.g., the traditional
Real-Time Streaming Protocol (RTSP), Real-Time
Transport Protocol (RTP) and Real-Time Transport
Control Protocol (RTCP). Increasingly, more
enterprises are using Web- and HTTP-based video
streaming via systems such as Microsoft Silverlight,
Apple QuickTime and Adobe Flash using protocols such
as HTTP Live Streaming (HLS), HTTP Dynamic
Streaming or Smooth Streaming.
NOTE 3
THE WAN CONNECTION IS CRITICAL TO
FAST STORAGE REPLICATION
Off-site storage backup often requires very large data
volumes to be transferred within a limited time window
to minimize the time the application is out of service.
This may be done at a set time during each day or
several times during a day to reduce the risk of data
loss. A full asynchronous database backup may involve
22/08/2013 12:27
How to Pick the Right WAN Optimization Solution for Your ...
Use Case
Key Objective
Enable enterprise
mobility strategies
To optimize user experience of applications being
accessed via a mobile device
Enterprise disaster
recovery
http://www.gartner.com/technology/reprints.do?id=1-1FCG...
Key WOC Functions
To minimize the enterprise disaster recovery time
Asymmetric Web acceleration
Symmetric software-based
WAN optimization controller
(SoftWOC) client
Compression and deduplication
Network traffic management
several 100GB of data, which can take several hours of
backup time. The size of the WAN connection, as well
as latency of that connection, has a significant impact
on achievable throughput. Typically, the distance
between the primary and secondary data centers is
between 10 km and 200 km. The general backup
latency requirement is 50 ms or less.
Gartner client inquiries indicate that around 40% of
enterprises perform storage backup and replication
over a shared WAN (MPLS), while the rest use a
dedicated WAN (Ethernet). For backup performed over
a shared network, QoS and traffic management are
needed to ensure bandwidth allocation for the backup
traffic.
Source: Gartner (April 2013)
Return to Top
Analysis
Start the WOC Selection Process by Defining and Prioritizing Use Case
Objectives and Functional Requirements
The only way in which network managers can properly select a WAN optimization solution is by clearly
identifying their key objectives and the functional requirements needed to meet these objectives. We
evaluate four specific use cases and establish high-level functional requirements for each use case.
Return to Top
Reduce WAN Costs
Controlling WAN costs is a key focus for enterprises. One aspect that many network managers
investigate is using bandwidth reduction to avoid upgrading a WAN link. This is done via three different
functions: (1) caching, (2) a combination of compression and deduplication, or (3) optimized delivery of
streamed video content. Bandwidth reduction is about reducing the amount of data being transmitted
across the WAN. While WAN cost control is a typical driver, reducing the amount of transmitted data
also reduces the user's latency experience.
Return to Top
Enable Data Center Consolidation by Improving Application Response Times
Gartner clients frequently discuss application performance problems during inquiries. Often, these
discussions are brought up in connection with data center consolidation and centralized application
deployments. There can be several causes for application performance problems, which also means
there is no single solution. The functionalities typically used to solve these performance issues are
application acceleration and network traffic control.
Return to Top
Review Enterprise Mobility Strategies to Improve Efficiency of Mobile and Remote
Workers
Mobility and use of personal devices continue to see strong enterprise interest. According to "Hunting
and Harvesting in a Digital World: The 2013 CIO Agenda," enterprises are already migrating to laptops
and mobile devices as the key user devices. In combination with increased adoption of bring your own
device (BYOD), cloud computing and working outside the office LAN, this means that network managers
need to consider how to deliver business applications with acceptable performance to individual
employees. This can be done via two dedicated mobile solutions: Web acceleration or SoftWOC clients.
Return to Top
Minimize the Enterprise Disaster Recovery Time
Replicating storage data for disaster recovery has a fairly simple goal: Move data from primary storage
to secondary storage, and back, during disaster recovery as quickly and securely as possible to meet
the enterprise recovery time objective, such as less than one hour to recover a business process after
failure. The faster that the replication process can take place, the less impact and downtime there is on
the business process and for end users. WAN optimization is often used to assist in this replication via
dedicated functions for acceleration of storage replication.
Return to Top
Use Defined Functional Requirements to Shortlist WOCs
Having defined specific functional requirements, the next step is to identify specific features and
capabilities with vendors' products that can best meet the requirements.
Return to Top
Multiple Ways to Reduce WAN Bandwidth and WAN Costs
Use Caching
Caching has been used for many years and is still supported in various manners by WOC vendors, such
as Blue Coat, Cisco Systems, Citrix, Exinda and Riverbed. However, the types of caching have increased
to meet the needs of current applications. Most WOC vendors support two types of caching:
File or object caching stores or copies entire files or objects locally, and thus avoids transmitting
the same file several times across the WAN. File caching is particularly useful for frequently
accessed files or objects that are more static in nature, and can reduce bandwidth use by more
than 90%. However, this is application-specific — for example, HTTP content, images, streaming
media objects or FTP files. Thus it's important to ensure not only that relevant applications are
supported, but also that the device provides sufficient storage. Leading devices support up to
2 of 5
22/08/2013 12:27
How to Pick the Right WAN Optimization Solution for Your ...
http://www.gartner.com/technology/reprints.do?id=1-1FCG...
15TB of disk storage capacity. More vendors are planning to offer video caching functionality.
Vendors should also offer content refresh and content distribution.
Byte caching identifies frequently repeated bit or byte patterns across all content that is
transmitted, and stores these byte patterns locally in a segment store. A key difference between
the two types of file caching is that byte caching is done without regard to the relationship
between file and objects, and thus caches byte patterns that appear in multiple files and file types.
Byte caching also works on more dynamic files — e.g., editing one word in a document removes
the benefit of file caching, but leverages the byte cache. Byte caching and segment store are
integrated components of compression and deduplication.
Use Compression and Deduplication
Compression and deduplication are supported by most vendors, including Blue Coat, Cisco, Exinda,
Riverbed, Citrix, Ipanema and F5. Typically, these are implemented as an integrated two-stage process
with a generic first-pass compression based on standard gzip, Lempel-Ziv (LZ) or similar algorithm, and
a segment-store second pass based on a proprietary deduplication algorithm (see Note 1). To achieve
the greatest reduction in data transmission, both processes must be employed.
The performance of vendor-specific solutions is highly dependent on algorithms and actual
implementations, but also depends on processing power and amount of memory supported by the
appliance. Today, vendors usually use solid-state drive (SSD) flash memory for higher data throughput,
and typically in the order of gigabytes to terabytes for high compression levels.
Network managers should evaluate solutions in the context of enterprise-specific data content before
making any assumptions about compression rates. Text can typically obtain a high compression rate,
while graphic (picture) files (GIF, JPG, etc.), music files (MP3, WMA, etc.), video files (AVI, MPG, etc.)
and other already compressed traffic achieve limited additional compression. It is not possible to
compress encrypted traffic, and, thus, it is essential that the WOC support Secure Sockets Layer (SSL).
As a general rule, data deduplication can reduce bandwidth needs by at least 60%. However, in many
cases, bandwidth may be reduced by 80% to 90% or even more.
Use Optimized Video Content Delivery
Enterprise adoption of video continues to grow, with a predicted increase to 16 hours per employee per
month by 2016, from 4.1 hours in 2012. This means that enterprises are interested in how to control
and reduce the impact on the enterprise WAN (see Note 2). This adoption is for real-time video
communications and streaming video, which is used for many different purposes, such as on-demand
training and education, live video streaming of corporate events and live executive briefings.
WOC vendors, such as Blue Coat, Cisco Systems, Citrix and Riverbed, are adding functionality to their
appliances that better manage the network impact of high volumes of video traffic. In addition to
traditional functionality for traffic quality of service (QoS), caching and deduplication, vendors are
increasingly adding media servers to optimize video performance via stream splitting, video
prepopulation and video caching. These techniques can reduce WAN bandwidth utilization by more than
90%.
Return to Top
Enable Data Center Consolidation by Application Acceleration
The typical source of application performance problems continues to be transport and application
protocols that are not designed for high-latency networks (e.g., for one-way latencies above 50
millisecond [ms] performance degradation). Acceleration technologies can reduce the latency impacting
end users by 60% or more.
Most vendors perform TCP optimization, which essentially optimizes all applications. As a general
guideline, these techniques can reduce latency impact up to 30% to 40%, depending on the
implementation and the amount of memory available for TCP windowing. When combined with
application layer latency compensation techniques, latency can be reduced by at least 60%.
In addition to generic transport protocol optimization, network managers should look for specific
application layer optimization, such as:
Application protocol optimization, as well as content prepositioning, pipelining, read-head, writebehind for Messaging API (MAPI), Common Internet File System (CIFS), Server Message Block
(SMB), Network File System (NFS), HTTP, etc.
Virtual Desktop Infrastructure (VDI) acceleration, QoS and deduplication/compression of Citrix's
Independent Computing Architecture (ICA), Microsoft Remote Desktop Protocol (RDP), VMware via
Microsoft Remote Desktop Protocol (RDP), VMware PCoIP
The result is significant acceleration of the applications. Expect 60% to 70% latency improvements in
addition to the TCP improvement. Vendors include Riverbed, Blue Coat, Exinda and Cisco Systems.
Return to Top
Improve Network Performance With Advanced Network Traffic Management
As enterprise bandwidth utilization continues to increase by over 30% annually, network managers
often struggle to maintain consistently good performance for their business applications. The issues are
related not only to lack of bandwidth planning, but also to use of recreational traffic, such as viewing
YouTube and seasonal sporting events online, which create unpredictable traffic patterns and traffic
congestion. By deploying application-level bandwidth usage monitoring and combining it with
application-specific traffic management, network managers can deliver consistent application
performance for critical applications, while maintaining a high level of network utilization — up to 80%.
This area includes a range of traffic management capabilities, such as:
Identify specific application traffic based on Internet Protocol (IP) address, port number,
application-specific identification, URL and HTTP
Display bandwidth utilization for each application broken down into WAN access or subnet, hosts
and user
3 of 5
22/08/2013 12:27
How to Pick the Right WAN Optimization Solution for Your ...
http://www.gartner.com/technology/reprints.do?id=1-1FCG...
Classify applications and URL/HTTP content into priority classes and priority queuing, and tag
applications for transport over the WAN
Allocate guaranteed bandwidth or performance for specific applications, as well as manage
bandwidth usage for all other applications
Throttle/shape or block/allow traffic by application, ports, protocols, URLs, URL category, users,
user groups, IP address, file type, Multipurpose Internet Messaging Extensions (MIME) type or
time of day
Vendors include Blue Coat, Exinda, Ipanema, Riverbed and Silver Peak. Network managers need to
differentiate between vendors that only do priority tagging of applications and queue management, and
vendors (such as Blue Coat and Exinda) that also perform active traffic control, such as shaping,
throttling and blocking based on a range of configurable conditions.
Return to Top
Support Individual Personal Devices
There are two methods available to improve application performance for mobile users:
An asymmetric solution is based on functionality in the current browser, and thus does not use
a dedicated software client on the user device. All Web browsers support negotiation of
compression on HTTP streams, typically using gzip and caching. In addition, Riverbed uses
JavaScripts for Web page optimization, such as Cascading Style Sheets, images and merging Web
page files for fewer roundtrips. This can reduce data volume by close to 50%, and reduce the
impact of latency by up to 60%. Vendors include Blue Coat, F5, Radware and Riverbed.
A symmetric solution is based on a SoftWOC client to be loaded onto the user device. A
SoftWOC offers features similar to a branch office appliance, across all applications, not just
browser-based applications, reducing latency up to 70%. These clients also support deduplication
and can reduce data volume up to 90%. In addition to understanding specific feature support,
network managers should evaluate support for OSs such as Windows CE, Windows XP, Windows
7, Windows 8, Apple iOS, Mac OS and Android. It is critical to evaluate mobile device
requirements for memory and processing power to perform application acceleration and
deduplication. Vendors that offer a SoftWOC include Riverbed, Cisco Systems, Circadence and
Citrix.
Return to Top
Reduce the Enterprise Disaster Recovery Time With Date Center to Date Center Storage
Replication Acceleration
The purpose of WAN optimization in this context is to speed up the transport of high data volumes
between data centers via data deduplication, IP/TCP acceleration and storage protocol acceleration for
the relevant storage replication solutions from Dell, EMC, Hitachi Data Systems, NetApp, Oracle and
others. While storage systems already perform block deduplication, WOCs can add byte-level
deduplication.
A key selection criterion is the data throughput the device supports. Therefore, network and storage
managers should start by defining their throughput objectives (e.g., 300GB per hour). Vendors'
marketing materials often focus on the speed of network interfaces as a key benchmark for the
throughput that their appliances support. Currently, the largest interface available is up to 10 Gbps (see
Note 3). However, the amount and type of memory supported by the device is equally, if not more,
important than the throughput. Strong data reduction is paramount to fast replication (and thus
throughput), and byte deduplication requires a large local store of very fast memory. Leading devices
support up to 5TB of SSD. Vendors include Silver Peak, Riverbed, F5 and Citrix.
Return to Top
Always Perform an On-Site Proof of Concept Before Signing the Purchasing
Contract
After compiling the vendor shortlist, network managers should test their preferred device before final
purchase. The purpose is twofold:
Network managers need to ensure that the device fulfills the specific functional needs of their use
cases within the enterprise network environment.
The enterprise networking staff needs to understand how easy or difficult it is to deploy and
operate the device.
There is a significant difference in how various vendors' devices are configured, deployed and managed.
Therefore, network managers need to include this information in their final selection criteria.
Testing is preferably done at real-world enterprise locations that provide a representative sampling of
the enterprise network (connection type, application and content traffic types, latency and bandwidth
characteristics). Alternatively, testing can be done on a test bed with a network impairment simulator
to mimic typical and worst-case network conditions. Network managers first should establish a baseline
noting their current application's performance and bandwidth use. Many WOCs monitor and report
these parameters. During initial testing, the network team should find out if the device can report the
impact on performance.
The test should be performed over a few weeks, typically three to five, to ensure most workload
conditions are tested. It is important to be prepared to fine-tune the device, as well as the applications
and changes to the network, to gain optimum performance. For example, the testers may have to turn
off application-based encryption and/or compression, use a particular port number with the application,
change QoS settings in the router or increase WAN bandwidth for a certain class of end users.
It is important that the enterprise network staff do not work in silos. They should work alongside the
vendor or vendor partner during the test phase to get hands-on experience. They should include
security staff in the proof of concept to understand and test the support of SSL, and the handling of
4 of 5
22/08/2013 12:27
How to Pick the Right WAN Optimization Solution for Your ...
http://www.gartner.com/technology/reprints.do?id=1-1FCG...
encryption keys and certificates. Finally, the enterprise network staff needs to report to the selection
team about the complexity and time involved in operating the device or devices.
Return to Top
© 2013 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered trademark of Gartner, Inc. or its affiliates. This publication may not be reproduced
or distributed in any form without Gartner’s prior written permission. If you are authorized to access this publication, your use of it is subject to the Usage Guidelines for
Gartner Services posted on gartner.com. The information contained in this publication has been obtained from sources believed to be reliable. Gartner disclaims all
warranties as to the accuracy, completeness or adequacy of such information and shall have no liability for errors, omissions or inadequacies in such information. This
publication consists of the opinions of Gartner’s research organization and should not be construed as statements of fact. The opinions expressed herein are subject to
change without notice. Although Gartner research may include a discussion of related legal issues, Gartner does not provide legal advice or services and its research
should not be construed or used as such. Gartner is a public company, and its shareholders may include firms and funds that have financial interests in entities covered
in Gartner research. Gartner’s Board of Directors may include senior managers of these firms or funds. Gartner research is produced independently by its research
organization without input or influence from these firms, funds or their managers. For further information on the independence and integrity of Gartner research, see
“Guiding Principles on Independence and Objectivity.”
About Gartner | Careers | Newsroom | Policies | Site Index | IT Glossary | Contact Gartner
5 of 5
22/08/2013 12:27