Mercury Interactive Identifies Four Most Common Causes of Web Site Performance Problems

Company Finds That 35% of All Performance Bottlenecks Can Only Be Found By

Testing From Outside the Firewall



Apr 17, 2001, 01:00 ET from Mercury Interactive Corporation

    SUNNYVALE, Calif., April 17 /PRNewswire/ -- Mercury Interactive Corp.
 (Nasdaq:   MERQ), the leading provider of enterprise testing and application
 performance management solutions, today published knowledge gleaned from
 thousands of load tests conducted with ActiveTest(TM), the company's hosted
 Web-based load testing service. Mercury Interactive's testing experts have
 traced the majority of Web site performance problems to the following sources:
 database tuning, network bottlenecks, and application server and Web server
 configuration. In addition, test results have revealed that 35 percent of
 performance bottlenecks occur outside the firewall -- due to tuning and
 configuration issues with routers, gateways and switches, as well as bandwidth
 constraints and ISP peering point problems. Performance problems that lie
 outside the firewall can only be found by testing Web sites via the Internet.
     The company's findings confirm that performance problems can exist in
 virtually any part of the Web infrastructure. The components used to build Web
 sites -- such as application servers, Web servers, databases, and caching
 appliances -- can be individually robust and scalable, but may not perform as
 expected when they are combined to run complex, customized Web applications.
 In fact, results collected by Mercury Interactive show that performance
 problems occur 98 percent of the time simply because the infrastructure
 components have not been tuned or configured properly for a specific
 application. Therefore, it is important to test and tune the entire system
 before going live, or before introducing major changes, to ensure optimal
 performance.
     "Mercury Interactive's ActiveTest allowed Fortune.com to examine the
 behavior of its redesigned Web site, under heavy and sustained virtual user
 traffic loads, without the hassles and cost of setting up an in-house testing
 lab," said Eric Porres, director of production in the FORTUNE Group at Time,
 Inc. "Identifying and resolving our Internet application's performance issues
 before going live helped us deliver the positive user experience our customers
 expect."
     "Mercury Interactive's ActiveTest experts have gained valuable knowledge
 by performing thousands of load tests on a wide range of Web applications,"
 said David Gehringer, senior product marketing manager for Mercury
 Interactive. "Our experts use this knowledge to help ActiveTest customers
 deliver Web sites that provide the best possible performance -- without adding
 costly hardware and software. In short, we help our customers cut their
 overall IT costs by helping them deliver maximum performance from their
 existing resources."
     The company's white paper, "Common Web Site Problems Identified by
 ActiveTest," outlines the most common Web site performance problems
 encountered during load testing with ActiveTest, and is accessible at
 http://www.mercuryinteractive.com/products/activetest/whitepapers/ .
 
     Problem Diagnosis: 'It's Like Peeling an Onion'
     Mercury Interactive's ActiveTest experts develop and run tests, from the
 company's nine load server farms, located at Tier 1 ISPs worldwide, to measure
 site performance by emulating Web traffic of hundreds, thousands, or even
 millions of individual users. During the load test, ActiveTest provides
 real-time online graphs to measure response times for transactions, and
 displays metrics on all components throughout the application and network
 infrastructure. With this data, ActiveTest experts, in conjunction with the
 customer, can quickly identify performance bottlenecks, pinpoint their causes
 and then fine-tune the performance of the sites to achieve the best possible
 user experience.
     "Oftentimes, solving a performance problem with the one component will
 reveal previously hidden problems with another component," said Gehringer.
 "Proper load testing is an iterative process. It's like peeling an onion and
 examining each layer looking for ways to improve performance. By taking this
 approach, our ActiveTest experts have typically helped our customers increase
 the performance of their Web applications by more than 400 percent -- without
 adding more hardware."
 
     Four Common Sources of Web Site Performance Problems
     The failure to scale to accommodate more traffic can be frequently
 attributed to poor software and hardware configurations. Oftentimes a company
 is in the process of implementing a new Web site design and does not have the
 data to configure or tune the various infrastructure components for a specific
 application. In helping companies overcome poor Web site performance or
 scalability, Mercury Interactive has most often traced the problems to the
 following sources:
 
     -- Database tuning -- Dynamic Web sites that rely heavily on database
        operations are particularly at risk for performance problems. As a
        result, nearly 27 percent of Web site performance problems can be
        traced to malfunctioning database operations, which are usually caused
        by insufficient indexing, fragmented databases, out-of-date statistics,
        or faulty application design.
     -- Network bottlenecks -- The ActiveTest team has found that in almost
        25 percent of all cases, the available network bandwidth to the
        Internet is insufficient to handle the desired load and causes delays
        in incoming and outgoing requests. Moreover, the ActiveTest team has
        frequently uncovered bottlenecks between the customer's Web site and
        the ISP, and has also found that improper tuning of firewall, load
        balancers, gateways and routers can contribute significantly to
        performance problems.
     -- Application Server configuration-Application server configuration
        issues, such as improper application configuration settings, and poor
        cache management or session management, account for more than
        23 percent of system bottlenecks. Since the application server is the
        hub of many complex system architectures, its performance problems can
        be difficult to identify and can impact the entire application. For
        example, application servers communicate directly with databases, so
        bottlenecks often are confused with database server transactions.
     -- Web Server configuration-Web server bottlenecks are the cause of poor
        performance in approximately 20 percent of all test cases. These
        bottlenecks are usually the result of improper server configurations
        and insufficient memory resources. For example, combining Web and
        application servers on the same CPU can quickly exhaust server
        resources.
 
     About ActiveTest
     Mercury Interactive's ActiveTest, powered by the industry standard load
 testing tool LoadRunner(R), is a hosted Web-based load testing service.
 ActiveTest provides full-scale stress testing of Web sites to identify
 capacity constraints and performance bottlenecks, and to pinpoint their cause.
 Since introducing ActiveTest in early 1999, Mercury Interactive has conducted
 thousands of hosted load tests -- the largest with 1.35 million concurrent
 connections -- and has gained substantial knowledge about how to deliver high
 performance Web sites.
 
     Pricing and Availability
     Mercury Interactive's ActiveTest is currently available with pricing
 starting at $15,000 (US) for a single scheduled load test. A series of six
 ActiveTest hosted load tests starts at $25,500 (US). Organizations who would
 like more information or begin testing can find more information at
 http://activetest.mercuryinteractive.com .
     Organizations purchasing a series of six or twelve ActiveTest load test
 runs will automatically qualify for the "Double Performance Guarantee." This
 guarantee assures customers that they can double the maximum number of
 concurrent users that can access their Web site, or cut the site's average
 response time in half, or double the site's maximum server throughput within
 six hosted load tests conducted by Mercury Interactive's ActiveTest experts.
 
     Mercury Interactive Corporation
     Mercury Interactive is the leading provider of enterprise testing and
 application performance management solutions that help businesses deliver a
 positive user experience. Mercury Interactive solutions enable its customers
 to turn Web application performance, scalability and user experience into
 competitive advantage. The company's performance management products and
 hosted services are open and integrated to best test and monitor
 business-critical Web applications.
     Mercury Interactive is headquartered in Sunnyvale, California and has
 40 offices worldwide. The company can be reached at www.mercuryinteractive.com
 or by phone at U.S. 408-822-5200.
 
     NOTE:  LoadRunner is a registered trademark and ActiveTest is a trademark
 of Mercury Interactive Corporation, Sunnyvale, California. Product and company
 names are used herein for identification purposes only, and may be trademarks
 of their respective companies.
 
 

SOURCE Mercury Interactive Corporation
    SUNNYVALE, Calif., April 17 /PRNewswire/ -- Mercury Interactive Corp.
 (Nasdaq:   MERQ), the leading provider of enterprise testing and application
 performance management solutions, today published knowledge gleaned from
 thousands of load tests conducted with ActiveTest(TM), the company's hosted
 Web-based load testing service. Mercury Interactive's testing experts have
 traced the majority of Web site performance problems to the following sources:
 database tuning, network bottlenecks, and application server and Web server
 configuration. In addition, test results have revealed that 35 percent of
 performance bottlenecks occur outside the firewall -- due to tuning and
 configuration issues with routers, gateways and switches, as well as bandwidth
 constraints and ISP peering point problems. Performance problems that lie
 outside the firewall can only be found by testing Web sites via the Internet.
     The company's findings confirm that performance problems can exist in
 virtually any part of the Web infrastructure. The components used to build Web
 sites -- such as application servers, Web servers, databases, and caching
 appliances -- can be individually robust and scalable, but may not perform as
 expected when they are combined to run complex, customized Web applications.
 In fact, results collected by Mercury Interactive show that performance
 problems occur 98 percent of the time simply because the infrastructure
 components have not been tuned or configured properly for a specific
 application. Therefore, it is important to test and tune the entire system
 before going live, or before introducing major changes, to ensure optimal
 performance.
     "Mercury Interactive's ActiveTest allowed Fortune.com to examine the
 behavior of its redesigned Web site, under heavy and sustained virtual user
 traffic loads, without the hassles and cost of setting up an in-house testing
 lab," said Eric Porres, director of production in the FORTUNE Group at Time,
 Inc. "Identifying and resolving our Internet application's performance issues
 before going live helped us deliver the positive user experience our customers
 expect."
     "Mercury Interactive's ActiveTest experts have gained valuable knowledge
 by performing thousands of load tests on a wide range of Web applications,"
 said David Gehringer, senior product marketing manager for Mercury
 Interactive. "Our experts use this knowledge to help ActiveTest customers
 deliver Web sites that provide the best possible performance -- without adding
 costly hardware and software. In short, we help our customers cut their
 overall IT costs by helping them deliver maximum performance from their
 existing resources."
     The company's white paper, "Common Web Site Problems Identified by
 ActiveTest," outlines the most common Web site performance problems
 encountered during load testing with ActiveTest, and is accessible at
 http://www.mercuryinteractive.com/products/activetest/whitepapers/ .
 
     Problem Diagnosis: 'It's Like Peeling an Onion'
     Mercury Interactive's ActiveTest experts develop and run tests, from the
 company's nine load server farms, located at Tier 1 ISPs worldwide, to measure
 site performance by emulating Web traffic of hundreds, thousands, or even
 millions of individual users. During the load test, ActiveTest provides
 real-time online graphs to measure response times for transactions, and
 displays metrics on all components throughout the application and network
 infrastructure. With this data, ActiveTest experts, in conjunction with the
 customer, can quickly identify performance bottlenecks, pinpoint their causes
 and then fine-tune the performance of the sites to achieve the best possible
 user experience.
     "Oftentimes, solving a performance problem with the one component will
 reveal previously hidden problems with another component," said Gehringer.
 "Proper load testing is an iterative process. It's like peeling an onion and
 examining each layer looking for ways to improve performance. By taking this
 approach, our ActiveTest experts have typically helped our customers increase
 the performance of their Web applications by more than 400 percent -- without
 adding more hardware."
 
     Four Common Sources of Web Site Performance Problems
     The failure to scale to accommodate more traffic can be frequently
 attributed to poor software and hardware configurations. Oftentimes a company
 is in the process of implementing a new Web site design and does not have the
 data to configure or tune the various infrastructure components for a specific
 application. In helping companies overcome poor Web site performance or
 scalability, Mercury Interactive has most often traced the problems to the
 following sources:
 
     -- Database tuning -- Dynamic Web sites that rely heavily on database
        operations are particularly at risk for performance problems. As a
        result, nearly 27 percent of Web site performance problems can be
        traced to malfunctioning database operations, which are usually caused
        by insufficient indexing, fragmented databases, out-of-date statistics,
        or faulty application design.
     -- Network bottlenecks -- The ActiveTest team has found that in almost
        25 percent of all cases, the available network bandwidth to the
        Internet is insufficient to handle the desired load and causes delays
        in incoming and outgoing requests. Moreover, the ActiveTest team has
        frequently uncovered bottlenecks between the customer's Web site and
        the ISP, and has also found that improper tuning of firewall, load
        balancers, gateways and routers can contribute significantly to
        performance problems.
     -- Application Server configuration-Application server configuration
        issues, such as improper application configuration settings, and poor
        cache management or session management, account for more than
        23 percent of system bottlenecks. Since the application server is the
        hub of many complex system architectures, its performance problems can
        be difficult to identify and can impact the entire application. For
        example, application servers communicate directly with databases, so
        bottlenecks often are confused with database server transactions.
     -- Web Server configuration-Web server bottlenecks are the cause of poor
        performance in approximately 20 percent of all test cases. These
        bottlenecks are usually the result of improper server configurations
        and insufficient memory resources. For example, combining Web and
        application servers on the same CPU can quickly exhaust server
        resources.
 
     About ActiveTest
     Mercury Interactive's ActiveTest, powered by the industry standard load
 testing tool LoadRunner(R), is a hosted Web-based load testing service.
 ActiveTest provides full-scale stress testing of Web sites to identify
 capacity constraints and performance bottlenecks, and to pinpoint their cause.
 Since introducing ActiveTest in early 1999, Mercury Interactive has conducted
 thousands of hosted load tests -- the largest with 1.35 million concurrent
 connections -- and has gained substantial knowledge about how to deliver high
 performance Web sites.
 
     Pricing and Availability
     Mercury Interactive's ActiveTest is currently available with pricing
 starting at $15,000 (US) for a single scheduled load test. A series of six
 ActiveTest hosted load tests starts at $25,500 (US). Organizations who would
 like more information or begin testing can find more information at
 http://activetest.mercuryinteractive.com .
     Organizations purchasing a series of six or twelve ActiveTest load test
 runs will automatically qualify for the "Double Performance Guarantee." This
 guarantee assures customers that they can double the maximum number of
 concurrent users that can access their Web site, or cut the site's average
 response time in half, or double the site's maximum server throughput within
 six hosted load tests conducted by Mercury Interactive's ActiveTest experts.
 
     Mercury Interactive Corporation
     Mercury Interactive is the leading provider of enterprise testing and
 application performance management solutions that help businesses deliver a
 positive user experience. Mercury Interactive solutions enable its customers
 to turn Web application performance, scalability and user experience into
 competitive advantage. The company's performance management products and
 hosted services are open and integrated to best test and monitor
 business-critical Web applications.
     Mercury Interactive is headquartered in Sunnyvale, California and has
 40 offices worldwide. The company can be reached at www.mercuryinteractive.com
 or by phone at U.S. 408-822-5200.
 
     NOTE:  LoadRunner is a registered trademark and ActiveTest is a trademark
 of Mercury Interactive Corporation, Sunnyvale, California. Product and company
 names are used herein for identification purposes only, and may be trademarks
 of their respective companies.
 
 SOURCE  Mercury Interactive Corporation