Understanding WAN Simulation and Network Testing Tool ...

Understanding WAN Simulation and Network Testing Tool ... screenshot

Text-only Preview

  1. Understanding WAN Simulation and Network Testing Tool Alternatives Gartner RAS Core Research Note G00144685, Will Cappelli, Debra Curtis, 2 April 2007 R2280 04242008 Proactive use of WAN simulation and network testing technologies early in the development life cycle of new Web- based and composite applications can better predict “real world” production performance. Key Findings • New Web-based application architectural styles make it more likely that different application components will wind up on different ends of a WAN link and that the execution of the application’s logic will require a complex pattern of network link traversals. • Centralization of applications increases the likelihood that end users will experience WAN latency as transactions traverse continents and oceans during the course of their processing. • The testing and quality assurance phase of the application development life cycle has become more important, and there is a premium on any technology that can do more to reveal the real-world performance characteristics of applications being tested in a virtualized test bed. Recommendations • Use WAN simulation and network testing technologies to accurately predict how applications will perform once they are released on the production network. • Network engineering and capacity planners should get involved earlier in the application life cycle, and introduce WAN simulation and network testing to the quality assurance lab. ANALYSIS Although the prospect of new network traffic types, such as voice over IP (VoIP) and IP television (IPTV), have fueled a slow but steady growth in the demand for WAN simulation and network testing technology during the past two years, a recent acceleration in that growth has been sparked by a convergence of application architecture and testing trends. We, however, anticipate a significant acceleration in demand growth as a result of the following four trends: modular architectures, globalized applications, increasing importance of testing and quality assurance (QA) and virtualized test labs. Application Architecture Trends • Modular architectures. New Web-based architectural styles, such as service-oriented architecture (SOA), representational state transfer and Ajax, are encouraging the hyperfragmentation of application code. The more modular a given architecture, the more likely that different components will wind up on different ends of a WAN link and that the execution of the application’s logic will require a complex pattern of network link traversals.
  2. 2 • Globalized applications. Applications are being globalized • Tool to help stress test our network, simulating a high-volume with some businesses showing a preference for a single central load on a circuit or a device so that we can determine instance over multiple instance implementations of in-house- when/how it fails developed and off-the-shelf applications. This increases the likelihood that a significant percentage of an application’s Requests vary from testing an application in a lab, to testing network latency, as experienced by the end user, will be attributable to devices and application performance in a preproduction staging the need for transactions to traverse continents and oceans area, to testing the network or an application that’s already in during the course of their processing. production. Depending on the requirement and the group that is planning to use the tool, different product types may be appropriate. Testing and Quality Assurance Trends • Increasing importance of testing and QA. Historically, a focus Tool Taxonomy on time-to-market and short-term cost minimization meant that Three fundamentally distinct types of WAN simulation and network the testing and quality assurance phase was given short shrift or testing technologies address these requirements; each type has abandoned all together. However, during over the past four varying applicability to different target users in the IT organization. years, the testing and quality assurance phase of the application development life cycle has come to assume paramount WAN emulators for application testing – Tools of this type are importance, fueled by competitive and regulatory considerations. focused on testing the contribution that the WAN makes to the performance and end-to-end latency of an application transaction. • Virtualized test labs. In the past, testing environments were The WAN emulator appliance attaches to the LAN in the QA lab largely replications of production environments, and hence, it and impairs the application’s network traffic, making it appear to was relatively easy to pinpoint where a network traversal event the application as though it were communicating over a WAN. would affect an application’s performance (even if it were not so They are designed to be linked to an application testing bed and easy to anticipate how much that performance would be are often integrated with application testing and QA tools. They are affected). Now, however, testing environments are increasingly used in preproduction rollout stages to predict application built on top of virtualized platforms. Although these environments performance at remote offices and when considering the will certainly enable logic and control flow to be tested, the performance of alternative application-hosting models. They are industry still does not have sufficient experience to articulate also frequently used with WAN optimization products in a proof of rules for mapping the performance characteristics of these concept to predict the change in performance levels to remote virtualized environments to the performance characteristics of sites prior to actual purchase and deployment. the production environments they are meant to model. Hence, there is a premium on any technology that can do more to reveal The WAN characteristics (such as latency, jitter, errors, packet loss the real-world performance characteristics of applications being and duplicated or impaired packets) can be manually defined, or in tested in a virtualized test bed. some products, the tool can record the actual characteristics between locations in the global network. The WAN issues are Inquiries from Gartner clients certainly attest to growing interest. injected into the application traffic in the test lab to predict the The questions come from a wide variety of roles in the IT application’s remote performance characteristics across complex organization and cross such a varying mix of requirements that it is networks, prior to rollout into production. This type of tool is used difficult for any one type of tool to achieve prominence. Example by developers, network engineers and QA/test professionals, use cases that clients are asking about include: usually with the assistance of network experts. The representative • Network emulation tool for application developers to use to and market-share-leading product of this description is Shunra’s give them a sense of how their application will behave over a Virtual Enterprise network simulator. Adtran’s Atlas 550, Anue’s WAN or a congested network network emulators, Apposite’s Linktropy 4500 WAN emulator, • Network testing tool to test changes to network topology Candela’s LANforge Ice, InterWorking Labs’ Maxwell, Itheon’s before they are implemented in a production network Network Emulator, PacketStorm’s network emulators, Simena’s • Tool to emulate WAN connections to test performance of Network Emulator and the open-source Network Nightmare tool client/server applications, such as Outlook to Exchange, over a are other examples. WAN link prior to deployment Model-based analytical simulators – Tools of this type enable • Stress test tool that can simulate a full user load on a business users to build visual and statistical models of WAN topologies and application (for example, PeopleSoft), testing not just the Unix study analytically the impact that changing parameter values have on servers, but also the network and firewall, to see if application performance. The tools require a sophisticated, well- infrastructure upgrades are needed trained user to get the most value, but they are simple and • Tool that will generate traffic that simulates production inexpensive compared to the cost of replicating the entire production applications in a controlled network staging environment so network in a lab. To simplify the definition of the model, network that whenever network equipment is added, upgraded or device configuration files are imported and application transactions reconfigured, we can ensure the network reacts to traffic in a are captured through packet trace or agent software. Often called predictable way “what if” analysis, these tools are used by network engineers, © 2007 Gartner, Inc. and/or its Affiliates. All Rights Reserved. Reproduction and distribution of this publication in any form without prior written permission is forbidden. The information contained herein has been obtained from sources believed to be reliable. Gartner disclaims all warranties as to the accuracy, completeness or adequacy of such information. Although Gartner’s research may discuss legal issues related to the information technology business, Gartner does not provide legal advice or services and its research should not be construed or used as such. Gartner shall have no liability for errors, omissions or inadequacies in the information contained herein or for interpretations thereof. The opinions expressed herein are subject to change without notice.
  3. 3 network architects, network planners and QA/test professionals to The Future profile the application and examine how network behavior and The WAN emulators for application testing most directly provide application performance change in the model when new users or answers to the forces driving the new acceleration in demand for new applications are added or when different network technologies WAN simulation and network testing functionality, although all three with different latency characteristics are inserted. Model-based tools types of tools are being deployed in application QA settings. are most often used to examine high-impact, mission-critical Indeed, we anticipate that, during the next year to 18 months, applications and to judge the broader impact of planned vendors of the other types of tools will take some steps toward infrastructure changes or application restructuring. Opnet’s IT Guru integration with application testing tool suites and that the software is a representative and market-share-leading product of analytical modeling packages will add application data types to this description. Compuware’s Vantage Predictor is another their metamodels. In addition, we expect that vendors in one example. At the low end, Network Instruments Observer can category will expand their functionality to cover other categories in perform some limited “what if” analysis for a very reasonable price. an integrated suite-style solution. We have mentioned the vendors in their primary categories, but there are already some that have Network traffic generators – Tools of this type are focused on begun the migration to cover additional categories. testing network devices and the end-to-end performance of networks, within which WANs make up a significant percentage of Beyond the need for WAN simulation technologies, the the fabric. They work by generating actual network traffic load hyperfragmentation of application architectures signals the need for according to scripted protocol or application messages. They can a new culture that integrates the disciplines of application mimic existing network conditions or create high-traffic conditions development and network architecture design. Although there will in the safety of a controlled lab environment to assess always be design processes and testing routines that are purely performance and predict the impact of change. Tools of this type oriented toward network architecture and those that are purely are sometimes found in conjunction with network protocol analyzer oriented toward applications, there is a gray area in between that, products, but there are also more-sophisticated, specialized thanks to SOA and other Web-oriented architectural styles, is products that can generate hundreds of network and application growing rapidly. This will place an increased emphasis on tools that protocols, sending test messages to network devices to assess can be used in the test lab and then move to the production their performance characteristics under load and error conditions. environment, providing data that compares pre-deployment The traffic generator can be configured to generate specific baseline estimates and predictions to actual performance volumes of network or application traffic, including voice traffic to measurements taken from the live infrastructure where there are measure end-to-end voice quality across IP networks. other applications fighting over the same resources. This will require the ability for test data to be integrated with and used by IT The more-sophisticated (and expensive) products in this category are operations monitoring tools in the production environment. most often used by network equipment manufacturers and telecommunications carriers to load-test their network hardware What You Need to Know devices, as well as verify the devices’ responses to specific network Network engineers generally bear the brunt of blame when end- protocol messages. Enterprise network engineers can use traffic user application performance suffers. However, the problem is generators to stage new network devices in the lab, perform stress often a poorly designed application that doesn’t take into account testing, measure the response to protocol and application changes the impact of deployment on a global WAN. To prevent the “blame before they are implemented, and assess the impact on the game” after deployment, it would benefit the network engineer to infrastructure caused by a mix of applications. This type of tool get involved earlier in the application life cycle and champion the requires that the actual network devices run in the lab or in production use of WAN simulation and network testing technologies to to receive the simulated network and application traffic. Example successfully deploy new networked applications with accurate products in this category include Candela’s LANforge Fire, ClearSight business service-level expectations for performance and availability. Packet Generator, Fluke Networks’ OptiView Traffic Generation, Ixia IxChariot, Network Instruments Observer Traffic Generator, Simena’s Traffic Generator, SolarWinds WAN Killer and Spirent’s SmartBits. InterWorking Labs’ Maxwell product also has some capabilities in this category, although it is primarily in the WAN emulation category.