Benchmark Strategies for Appliation Acceleration
Description
The authoritative, unbiased source for IT WHITE PAPERAPERcertification, research and testing
May 2006
BenchmarkingA white paper
commissioned by
Akamai Technologies, Strategies fforCitrix Systems,
Foundry Networks,
Radware and Application
Streamcore Systems
Document #206114 Acceleration
©2006 The Tolly Group Benchmarking Strategies for
Application Acceleration
Terms oof UUsage
Entire contents © 2006 The Tolly Group, Inc. All rights reserved.
USE THIS DOCUMENT ONLY IF YOU AGREE TO THE TERMS LISTED HEREIN.
This document is provided, free-of-charge, to help you understand whether
a given product, technology or service merits additional investigation for
your particular needs. Any decision to purchase must be based on your own
assessment of suitability. This evaluation was focused on illustrating specific
features and/or performance of the product(s) and was conducted under
controlled, laboratory conditions and certain tests may have been tailored to
reflect performance under ideal conditions; performance may vary under
real-world conditions. Users should run tests based on their own real-world
scenarios to validate performance for their own networks. Commercially
reasonable efforts were made to ensure the accuracy of the data contained
herein but errors can occur.
The test/audit documented herein may also rely on various test tools the
accuracy of which is beyond our control. Furthermore, the document relies
on certain representations by the sponsor that are beyond ...
May 2006
BenchmarkingA white paper
commissioned by
Akamai Technologies, Strategies fforCitrix Systems,
Foundry Networks,
Radware and Application
Streamcore Systems
Document #206114 Acceleration
©2006 The Tolly Group Benchmarking Strategies for
Application Acceleration
Terms oof UUsage
Entire contents © 2006 The Tolly Group, Inc. All rights reserved.
USE THIS DOCUMENT ONLY IF YOU AGREE TO THE TERMS LISTED HEREIN.
This document is provided, free-of-charge, to help you understand whether
a given product, technology or service merits additional investigation for
your particular needs. Any decision to purchase must be based on your own
assessment of suitability. This evaluation was focused on illustrating specific
features and/or performance of the product(s) and was conducted under
controlled, laboratory conditions and certain tests may have been tailored to
reflect performance under ideal conditions; performance may vary under
real-world conditions. Users should run tests based on their own real-world
scenarios to validate performance for their own networks. Commercially
reasonable efforts were made to ensure the accuracy of the data contained
herein but errors can occur.
The test/audit documented herein may also rely on various test tools the
accuracy of which is beyond our control. Furthermore, the document relies
on certain representations by the sponsor that are beyond ...
Sujets
Informations
| Publié par | Besif |
| Nombre de lectures | 61 |
| Langue | English |
Extrait
certification, research and testing
for
May 2006 A white paper commissioned by Akamai Technologies, Citrix Systems, Foundry Networks, Radware and Streamcore Systems Document #206114
©2006 The Tolly Group
IT
Benchmarking Strategies f or Application Acceleration
WHITE PAPER
Terms o f U sage
Entire contents © 2006 The Tolly Group, Inc. All rights reserved. USE THIS DOCUMENT ONLY IF YOU AGREE TO THE TERMS LISTED HEREIN. This document is provided, free-of-charge, to help you understand whether a given product, technology or service merits additional investigation for your particular needs. Any decision to purchase must be based on your own assessment of suitability. This evaluation was focused on illustrating specific features and/or performance of the product(s) and was conducted under controlled, laboratory conditions and certain tests may have been tailored to reflect performance under ideal conditions; performance may vary under real-world conditions. Users should run tests based on their own real-world scenarios to validate performance for their own networks. Commercially reasonable efforts were made to ensure the accuracy of the data contained herein but errors can occur. The test/audit documented herein may a lso rely on various test tools the accuracy of which is beyond our control. Furthermore, the document relies on certain representations by the sponsor that are beyond our control to verify. Among these is that the software/hardware tested is production or production track and is, or will be, available in equivalent or better form to commercial customers. The Tolly Group provides a fee-based service to assist users in understanding the applicability of a given test scenario to their specific needs. Contact us for information. When foreign translations exist, the English document is considered authoritative. To assure accuracy, only use documents downloaded directly from The Tolly Group's Web site.
May 2006
©2006 The Tolly Group
Tolly G roup V endor S ervice
With more than 17 years experience validating leading-edge Information Technology products and services; The Tolly Group has built a global reputation for producing accurate and unbiased evaluations and analysis. We employ time-proven test methodologies and fair testing principles to benchmark products and services with the highest degree of accuracy. Our "Up-to-Spec" service provides the custom testing complement to the "standard", granular tests offered in "Tolly Verified". See our "Up-to-Spec"This document HomePage . was authored b : Plus, unlike narrowly focused testing labs, TheTo l yGroup combines its vast y technology knowledge with focused marketing services to help clients better Charles Bruno, position product benchmarks for maximum exposure. Executive Editor The Tolly Group
noitatrategies forAppilacitnoA cclereeBmhcnikraS gn
Table of Contents 6 About T ollyEdge 6 Acknowledgements 6 Overview 6 Wide-a rea a pplications: A p roblem i n n eed o f a s olution 7 Step 1 : C lassify t he P roblem 8 Step 2 : U nderstand t he B asics o f A pplication A cceleration 9 On C aching 9 On C ompression 9 On L oad B alancing 10 On P rotocol O ptimization 10 On Q uality o f S ervice 10 On R oute O ptimization 10 Akamai's T op 1 0 T ips f or A pplication A cceleration S uccess 10 Step 3 : E stablish S ervice L evels 12 Step 4 : D econstruct t he A pplication P athway 12 Step 5 : U nderstand y our P lanning H orizon 13 Step 6 : P erform a C ost/Benefit A nalysis 13 Step 7 : C reate a B enchmark P rototype 14 Citrix M aps a P athway t o S uccess
ies forA StrategnoA ccleppilacit upy Malyolro Gramhgnik6002cneBehT 60T ©02
14 Step 8 : U nderstand t he E lements L imiting P erformance 14 Server s ide 14 Bottleneck: Single-server limitations and over-utilized servers 14 Solution: Server load balancing 14 Bottleneck: SSL processing 15 Solution: SSL offload 15 Client s ide 15 Bottleneck: Multiple fetches of the same data 15 Solution: Local site cache 15 Solution: Smart cache in client 16 Solution: Compression 16 Network T ransport 16 Bottleneck: Insufficient bandwidth and/or high link latency 16 Solution: Reduce data traversing the link via data compression at the edge 17 Solution 2: Reduce data traversing the link via caching 17 Solution 3: Protocol optimization 17 Bottleneck: Too many applications fighting for bandwidth 17 Solution: Quality of Service and/or traffic shaping 17 Bottleneck: Clogged pathways leading to traffic delays 17 Solution: Route optimization 18 Radware S erves U p A pplication A cceleration C hecklist
retaoin Aonelcclipptica seiArofrtS geta Glyol Ty MauprocneB6002gnikramh©T eh0260
22
21
Quality o f S ervice
Step 1 0: M easure y our s uccess
23
20
STREAMCORE c ommon s ense a pplied t o a pplication a cceleration
21
Protocol O ptimization
18
Compression
19
Load b alancing
Step 9 : T est-d rive t he T echnology
May 2006
Caching
18
no
©2006 The Tolly Group
aritecelBhcnekram
©2006 The Tolly Group
But as more users reach out to access a growing number of appli-cations over the public Internet from all corners of the globe, enter-prise networks are beginning to suffer blows to performance and availability. Add to this the melding of voice, real-time video and data into these applications and you can clearly see why users might experience slow application response times, timed-out e-commerce sessions or failures to fulfill database requests. These neg-ative interactions with online applications can result in low adoption rates of applications that promised to be productivity enhancers or losses in revenue for what were supposed to be new lines of online business. Both are disastrous outcomes for companies looking to improve user interactions and corporate revenue. 6 May 2006
Today, global companies, e-commerce sites and other organiza-tions are turning to Web-based applications and server-based com-puting to better connect with employees, customers and business partners. They hope to boost productivity for remote workers, enhance online transactions and allow workers to collaborate across great distances. They are "Webifying" legacy applications, as well as bringing onboard new Web-ready and server-based appli -cations to accomplish such tasks as payroll, ERP and CRM.
This report is the result of a series of interviews with major players in the application acceleration market in combination with The Tolly Group's own experts. We'd like to thank Akamai Technologies, Citrix Systems, Foundry Networks, Radware and Streamcore Systems. Over view Wide-a rea a pplications: A p roblem i n n eed of a s olution
TollyEdge gives IT executives unparalleled insight into the issues under-scoring emerging technologies. Drawing upon the extensive expertise of The Tolly Group, Inc. and participating sponsors, you gain the knowledge necessary to develop a strategic and successful technology rollout. Acknowledgements
Benchmarking S trategies f or Application A cceleration
About TollyEdge
attr Sngfos ieegacilppArccA noitatioelernnchmarkiBe
7
May 2006
By eliminating or reducing the chokepoint or chokepoints in the local area, middle mile and client side, you allow a business transaction to take place more quickly and inevitably more efficiently. Reducing the wait time and the overall elapsed time for the transaction allows a user
If implemented at the correct locations in the application pathway, application acceleration technologies will let you extract more value from your enterprise network end-to-end than you thought possible.
In many areas of technology, knowing your goal — in this case, speed-ing application response time - helps you narrow down the product field you need to consider. Not so in the application acceleration market. All products in this arena — which cut a wide swath — have that same goal. Instead, you must hone in on your exact technological problem to find the appropriate solution.
Step 1: Classify the Problem
Normally, you hear people say "try before you buy." We believe when it comes to application acceleration, you first must under-stand what you're trying.
In this paper, we introduce you to the issues involved in wide-area application slowdowns. You'll learn where they occur in the network and how you can unclog those chokepoints whether you are dealing with office-to-office or widely distributed applications. With 10 simple steps, you'll have the knowledge you need about the technology-driven products on the market and how they match up to your network needs — allowing you to make intelligent, strategic purchasing decisions.
©2006 The Tolly Group
Instead, users should consider application acceleration technolo-gies — a growing and diverse market of products cloaked in mar-keting buzzwords that range from Web front-end devices to content delivery platforms to optimization appliances. Though they address the problem in different ways — some in the LAN or data center, some in the middle mile, and some at the client side — these prod-ucts all have the same goal: to speed up your application response time and therefore allow more business transactions per unit of time per remote user. After all, if you boost user productivity, you boost revenue.
The traditional response that IT groups have had to application per-formance and availability problems is to throw more servers and more bandwidth into the mix. This solution creates additional havoc as there is now more equipment to manage and wasted money on bandwidth that will be chewed up instantly by hungry applications.
Step 2: Understand the Basics of Application Acceleration
There is a natural byproduct that occurs when you address poor appli-cation response time – you unburden your network resources. By offload-ing tasks or sharing the load among multiple resources, you relieve the taxing duties on your heavy lifters in the network, gaining a better TCO from your investment. Optimizing the strengths of individual resources as well as the middle mile allows you to get more done in a more efficient manner.
With Web-based computing, very few resources are needed for applica-tions with static data, but if graphics-intensive applications are run this way, then they can generate a significant traffic load and increase the demands on the wide-area link.
There are generally two ways that applications can be delivered over the wide area to users — Web-based or server-based. Server-based computing differs from Web-based applications in that the actual application exe-cutes on the server and only the screen data is sent across the WAN.
or customer to achieve more in a shorter period of time. In many cases, this allows you to service more users and expand your business.
May 2006
We'll dig into each of these areas in more depth later in the paper, but understand there are three categories for application acceleration solutions: symmetric (something you need at both ends of a connection), asymmet-ric (something you need at one end), and those offered as a service.
©2006 The Tolly Group
8
Any deterioration in performance should be examined in three parts: the first mile or server side; the middle mile or wide-area link; and the last mile or client side.
Both approaches share common acceleration techniques, which we'll share with you in this paper. But before you can understand how to speed your applications, you must first understand the possible impedi-ments that adversely affect performance.
Depending on the nature of the application, one element may be more critical than another. Chances are, though, that response times could be improved by optimizing all three areas.
On the server side, you have time spent processing requests and queu-ing them for transmission. In the middle mile, there is time spent moving the resulting data packets across the network to the intended location. And in the last mile, time is consumed post-processing and rendering the data on the client's screen.
lecctareitacA nonioikgnS rteBcnmhraforAppliategies 
On C aching
Within these categories, there are numerous methods used to acceler-ate applications: caching, compression, load balancing, Q uality of S ervice, route optimization and protocol optimization. Some of these methods are location-specific and are applied in one part of the appli-cation path. Others can be applied in two or more places along the path between the client and server.
On L oad B alancing
Compression reduces the amount of time it takes for a given piece of data to traverse the WAN. The process involves compressing the data before it is fed into the wide-area link and decompressing it — or reconstituting the information into its original size — at its destina-tion. This minimization allows the WAN link, from the application point of view, to send the data at a higher rate than the physical link. For instance, compression can lead a T-1 link (1.5 Mbps ) with compression technology enabled at both ends to transmit 3 Mbps, 5 Mbps or more depending on the nature of the data being transmit-ted. Compression can be symmetric (with devices or software at both ends) or asymmetric (with a device or software at one end).
On C ompression
Caching can occur on the server side, the client side or the middle mile. Each has its benefits in increasing application response time. The aim of caching is to put frequently requested content close to the user so network resources aren't drained on repetition. For instance, static elements of a Web page can be cached on a local server, leaving the network more resources to deal with the dynamic aspects of the page. This saves on server load and bandwidth consumption. Caching can also be beneficial for dynamic data that will be viewed by large numbers of users before changing, such as weather reports or news headlines. You can set expiration times that allow you to gain the performance benefits of caching even for data that changes within fractions of seconds.
May 2006
Load balancing refers to a data center device that intercepts inbound application requests and distributes them to a farm of servers. This ensures that the capacity of a single server does not become a chokepoint. There are two types of load balancing – global server load balancing and local server load balancing. Global server load balancing, coupled with caching, can deliver significant performance improvements as it delivers the nearest
©2006 The Tolly Group
9
Application acceleration testing tools (testing software, boxes and other products are available) should be used to meet service-level goals established by your organization. You must know what bench-marks you have to reach — for instance, how many transactions per second — to fulfill your revenue and other business requirements. To do this successfully, you need to classify your application data into different pools.
Akamai's T op 1 0 T ips f or Application A cceleration S uccess 1. Ask questions about the business importance of the transaction and application and the impact to the business if the application is slow or unavailable. 2. Audit help desk tickets for availability and end-user response time issues. 3. Be on the lookout for any seasonality and time-of-use issues that may impact end users. 4. Use an application performance management (APM) tool to benchmark your current state. Run the tool for a month to understand time-of-use and seasonality issues. 5. Get information to split overall trans-action times into client, network and server times. 6. Consider integrating an APM tool into a service level management tool to aunndd earvsatialanbdi lirtoy oits scuaeus.se of performance On R oute O ptimization 7.IIfn teexrtneent dfionrg caopsptl iscaavtiiongnss oorn teox ttehneding Finally, route optimization allows data to be sent over the most effi-applications to new geographies cient path through the network. This technique senses if network understand the costs to the busi,ness devices or routes between network devices are experiencing poor if Internet performance slows and performance and can reroute traffic accordingly, resulting in a adversely impacts end users access faster application response and data delivery time. the application.Take into consideration offload savings if /when using a managed This is just a high-level view of each area. We'll dig down and see service provider. what each feature does to improve application response time and 8. Ask your vendor how the prospective how you can select the ones that are right for you. application acceleration solution handles pmeirdfdolre mmainlce.e issues of the Internet,or Step 3: Establish Service Levels 9. Ask vendor if the application acceleration solution has a service level agreement (SLA) to guarantee performance and availability. 10. Ask the vendor how the application acceleration solution handles infra-structure and bandwidth on demand to increase availability and reduce end-user response times.
10
May 2006
On Q uality o f S ervice Quality of Service (QoS) deals with the relative importance assigned to various application data types. By default, networks treat all traf-fic equally, employing a best-efforts policy for sending packets across the WAN. However, when you employ QoS, you can prioritize traffic, based on the critical nature of your applications. For instance, VoIP and videoconferencing are latency-sensitive so you need to make sure those packets are transmitted differently than E-mail or file sharing. You can employ techniques that introduce as lit-tle latency as possible.
and best performing instance of an application. Local server load balancing shares the load of requests coming into a data center. On P rotocol O ptimization Protocol optimization refers to a set of functions that deal with the handoffs between two ends. This approach is best used for "chatty" protocols such as CIFS or TCP, which require a lot of back and forth that is time-wasting and go on below your application. Protocol optimization tools offload those tasks and help make the most of each connection.
©2006 The Tolly Group
For example, if you are supporting a radiology team that needs to send images to doctors for immediate surgical consultation, then you want to accelerate all traffic having to do with that imaging application. You probably don't care about plain-old Web traffic coming from that net-work. But if you're a research firm whose team depends on the Web for their analysis of issues, then speeding Web traffic and delivery of Web pages is your goal. Finally, if you're an e-commerce site that lets cus-tomers design their own jeans, you'll want access to the back-end database that fills in sizing forms and collects credit card information to be top-notch. Knowing what applications need attention is critical to alleviating your chokepoints.
Prioritizing and classifying your applications will make choosing the right tool that much easier.
As a final example, applications that deal with rich media, such as voice over IP, face latency problems — especially when you consider that these applications are flowing over costly WAN links. If they are the core of your revenue-generating business, then you need to configure your network to give them ample resources.
11
May 2006
The best way to prioritize your business transactions is to classify them. These classes will allow you to identify characteristics of the traffic that directly affect its performance. Take for example applications that require up-to-the-minute data like stock tickers. These put a heavy burden on database servers at the data center. They must contend with the delay inherent in the request/response dynamic.
Don't try to tackle all your applications at once. This will only lead to confusion. You must start with one application, solve its chokepoints and then go to the next. Each application has its own issues that need to be addressed individually. For instance, you'll have little luck trying to speed videoconferencing traffic and E-mail traffic simultaneously. Be honest about what applications lead to boosted productivity and increased revenue. All others can wait.
Batch update applications, such as those associated with banking or inventory for field sales teams, must contend with bandwidth and capacity limitations. As fat as some pipes are, they struggle to handle large file transfers. This can lead to unbearable download times.
Applications that deal with e-commerce have issues dealing with the security of sensitive information like credit card data. They typically use Secure Sockets Layer (SSL), which, while necessary, imposes a perform -ance burden when the packets are decrypted.
©2006 The Tolly Group
© 2010-2024 YouScribe