Inexpensive Web Hosting - J2EE Hosting Sun Java, Mysql5, Java, Jsp, Servlet Developing Blog

CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING The Buckle Zone The buckle zone for the Buy High, Sell Low application occurred at 700 users. At this point, all use cases exceeded their SLA average values for greater than 80 percent of requests. Degradation Model The aggregate use case response time degradation model is shown in Figure 9-7. The aggregate response time degradation model plots the average response time buffer percentage against the user load. Figure 9-7. The aggregate response time degradation model plots the average response time buffer against user load. In this case, the average response time buffer percentage hits zero at a user load of 550 users. The response time buffer follows nearly an exponential pattern and crosses SLA boundaries at 550 users, so once the SLA is violated, the system can only sustain 100 to 125 users until the application is deemed completely unusable by the users. Unusable is defined as response times that exceed their buffer by more than 50 percent. The environment is primarily bound by CPU utilization in the application server tier. Figure 9-8 displays an aggregate of all CPUs present in the application server tier, and in this figure, you can plainly see that CPU utilization is trending upward.The application server tier CPU aggregate, shown in Figure 9-8, illustrates that by 650 users, the CPU spikes at over 90 percent and then continues to increase, staying over 95 percent utilization at 725 users. The alarming component of Figure 9-8 is the near linear increase of CPU utilization to user load. A linear increase indicates that if the application cannot be refactored to reduce CPU utilization, then tuning efforts are always going to be battling CPU limitations.
Searching for affordable and proven webhost to host and run your servlet applications? Go to Linux Web Hosting services and you will find it.

244 CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING The SLA Violation Point The SLA violation point for the Buy High, Sell Low application occurred at 550 users. Table 9-5 reports a summary of the use case behavior at the SLA violation point. Table 9-5. Use Case Summary at the SLA Violation Point Use Case SLA Ave SLA Dist SLA Max Actual Ave Actual Dist Actual Max Actual SD Actual 2xSD Delta Resp Time Buffer Login 8 sec. 95% 13 sec. 7.8 sec. 92% 12 sec. 2.2 sec. 4.0 sec. Violation Add Stock 10 sec. 95% 14 sec. 9.5 sec. 95% 13.7 sec. 2.0 sec. 3.9 sec. 5% Hist Query 10 sec. 95% 14 sec. 8.9 sec. 96% 13 sec. 2.9 sec. 3.9 sec. 11% Hist Graph 12 sec. 95% 16 sec. 11.1 sec. 91% 15.8 sec. 2.8 sec. 3.2 sec. Violation Stock Disc 8 sec. 95% 12 sec. 7.4 sec. 95% 10 sec. 2.5 sec. 3.0 sec. 7.5% Profile Mgmt 12 sec. 95% 16 sec. 10.2 sec. 94% 15.7 sec. 2.5 sec. 5 sec. Violation Summary 50% Violation The Saturation Point The saturation point for the Buy High, Sell Low application occurred at 600 users. Table 9-6 reports a summary of the use case behavior at the saturation point. Table 9-6. Use Case Summary at the Saturation Point Use SLA SLA SLA Actual Actual Actual Actual Actual Delta Resp Case Ave Dist Max Ave Dist Max SD 2xSD Time Buffer Login 8 sec. 95% 13 sec. 12.8 sec. 22% 37 sec. 6.2 sec. 12.0 sec. Violation Add 10 sec. 95% 14 sec. 19.5 sec. 14% 32.7 sec. 8.0 sec. 12.9 sec. Violation Stock Hist 10 sec. 95% 14 sec. 18.9 sec. 34% 23 sec. 3.9 sec. 4.9 sec. Violation Query Hist 12 sec. 95% 16 sec. 17.1 sec. 33% 25.8 sec. 4.8 sec. 3.2 sec. Violation Graph Stock 8 sec. 95% 12 sec. 15.4 sec. 27% 18 sec. 6.5 sec. 3.0 sec. Violation Disc Profile 12 sec. 95% 16 sec. 20.2 sec. 42% 27.7 sec. 5.5 sec. 5 sec. Violation Mgmt Summary 100% Violation
Visit our web design programs services for an affordable and reliable webhost to suit all your needs.

CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING The test environment includes clustering with AppServer1 using AppServer3 as its secondary server, AppServer2 using AppServer4 as its secondary server, and vice versa. In this way, the environment is resilient not only to application server instance failure, but also to hardware failure. This configuration adds additional performance overhead, but it meets the predefined availability and failover requirements. Monitoring Configuration The monitoring employed during this capacity assessment was configured to poll operating system, database, Web server, and application server statistics after every minute. The load tester was responsible for recording the overall request response times, while light bytecode instrumentation was employed to report tier-level response times. The bytecode instrumentation was not configured to record method-level statistics. Capacity Analysis This section presents the observations and conclusions derived in this capacity assessment. Expected Usage The expected usage for the Buy High, Sell Low application is 500 users. Table 9-4 reports a summary of the use case behavior at the expected load. Table 9-4. Use Case Summary at Expected Usage Use Case SLA Ave SLA Dist SLA Max Actual Ave Actual Dist Actual Max Actual SD Actual 2xSD Delta Resp Time Buffer Login 8 sec. 95% 13 sec. 6.2 sec. 97% 9.7 sec. 1.2 sec. 2.0 sec. 22.5% Add Stock 10 sec. 95% 14 sec. 8.5 sec. 97% 12 sec. 2.0 sec. 3.2 sec. 15% Hist Query 10 sec. 95% 14 sec. 8.2 sec. 96% 12 sec. 2.2 sec. 3.5 sec. 18% Hist Graph 12 sec. 95% 16 sec. 10.1 sec. 98% 14.2 sec. 1.8 sec. 2.2 sec. 15.8% Stock Disc 8 sec. 95% 12 sec. 6 sec. 99% 8.2 sec. 1.5 sec. 2.0 sec. 25% Profile Mgmt 12 sec. 95% 16 sec. 9.7 sec. 95.5% 15 sec. 2.5 sec. 5 sec. 19.16% Summary 19.24%
If you are looking for affordable and reliable webhost to host and run your business application visit our ftp web hosting services.

242 CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING with the following scenario distribution: Scenario Distribution Successful Login 94.5% Valid username, invalid password 5% Invalid username 0.5% Test Platform Topology The test platform consisted of six physical machines: Two Web servers Two application servers Two database servers Two application server instances run on each physical application server, totaling four application server instances. Figure 9-6 illustrates this topology. Figure 9-6. The test environment topology
Note: In case you are looking for affordable and reliable webhost to host and run your j2ee application check Vision J2ee Web Hosting services.

CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING The performance of the Buy High, Sell Low application has degraded with the release of version 2.0. The average response time degradation is 12 percent, and the average resource utilization at expected load has increased by 7 percent. Throughput at expected usage has likewise degraded by 10 percent. The maximum capacity has decreased from 650 users to 550 users, a degradation of 15.3 percent. I recommend additional hardware resources for addition to the environment while the source code is examined to identify the root of the performance degradation. Test Profile The capacity assessment was implemented using in-house load testing technology exercising the use cases shown in Table 9-3. Table 9-3. Test Profile Use Case Distribution Weight Login 0.1 Add Stock 0.1 Historical Query 0.2 Historical Graphing 0.2 Stock Discovery 0.2 Profile Management 0.2 The load test was configured to ramp up linearly over 30 minutes to the expected user load of 500 users. The test then implemented a graduated step sized at 25 users to ramp up over 5 minutes and hold for 5 minutes before initiating the next step. Test Script Configurations The following section summarizes the test script configurations. It includes detailed information about the primary use case scenario and summarizes the scenario distributions. Login The primary scenario for this use case is the successful login of a user with a valid username and a valid password. The steps for this scenario are summarized as follows: Request Think Time SLA Ave SLA SLA Flexibility Maximum GET /stock/index.html 10 sec. 3 sec. 95% 5 sec. POST /stock/login.do End 5 sec. 95% 8 sec.
We recommend you use shared web hosting services, because many users agree that it is cheap, reliable and customer-satisfying webhost.

240 CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING Sample Capacity Assessment Report Excerpts from the various sections of a Capacity Assessment Report follow. An actual Capacity Assessment Report may be 20 to 50 pages or more in length, so this sample attempts to reproduce each major section and include at least one major item in each section. You can fill in the remaining components with performance observations relevant to your environment. Executive Summary In this capacity assessment, the Acme Buy High, Sell Low stock application was evaluated in a mirrored production environment for performance. The expected user load for this application is 500 users, and the observations extracted from the test are illustrated in Figure 9-5. Figure 9-5. The Buy High, Sell Low environment s behavior as load increases Figure 9-5 can be summarized by the following observations: At the expected user load of 500, all use cases satisfy their SLAs. The first SLA violation is observed at 550 users. The environment s saturation point occurs at 600 users. The environment enters the buckle zone at 700 users. Use cases currently maintain an average response time buffer of 19.24 percent and based upon current trend analysis, this will dissipate rapidly over the next three months. My estimates suggest that the current environment will be in violation of its SLAs within five months. The test results indicate that the environment is CPU-bound and requires additional application server hardware to mitigate the five-month degradation point.
We recommend you use shared web hosting services, because many users agree that it is cheap, reliable and customer-satisfying webhost.

CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING If your organization is mature enough to routinely perform capacity assessments throughout the development of an application, then the impact analysis can be mostly automated, because it tracks performance differences between response times and resource utilizations for the same use cases and very similar, if not identical, test scripts. But if you are like most companies for whom performing a formal capacity assessment on each significant iteration is not feasible and who reserve capacity assessments for released code, then the task is a little more daunting and requires deep, domain-specific analysis. In this case, the summary of the impact analysis should be performed using response time buffer percentages. Recall that the response time buffers measure the percentage difference between the observed performance at a specific user load and the SLA. With this measurement, you can assess the performance of application functionality at specific user loads and determine whether a particular functional element degraded or improved in a subsequent release; you measure the degradation or improvement against the SLA defined for that functionality. If the SLA is renegotiated as a result of new or changed functionality, then an altered response time will not skew the impact analysis. The purpose of the impact analysis is to identify the following: General capacity impact of code changes, including the performance at the expected load, the SLA violation point, the resource saturation point, and the buckle zones Specific degradations and improvements of use cases Specific degradations and improvements in resource utilizations The sample Capacity Assessment Report later in this chapter provides additional details about the impact analysis. Analysis and Recommendations The analysis and recommendations section provides a conclusion to the Capacity Assessment Report. As such, it summarizes the findings again, but includes information about the impact of the findings on the business process and provides recommendations. It attempts to answer the following questions: What is the performance at the current or expected load? What load can the environment support and still satisfy SLAs? At what point does the environment need to be upgraded? What is the nature of that upgrade? Should it add more application server instances, or modify application server configurations (heap size, thread pools, connection pools, and so on)? At what point does the environment require additional hardware? If you have any insight into seasonal patterns, marketing promotions, or any other trending information that will affect user load, this information should be summarized or referenced here to justify your recommendations with forecasted behavior.
You want to have a cheap webhost for your apache application, then check apache web hosting services.

238 CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING Use case response times Utilization of each relevant resource Application throughput Each of these graphs should be overlaid with the following identified performance zones: Expected usage to SLA violation point SLA violation point to resource saturation point Resource saturation point to buckle zone The purpose of this section is to identify not only the behavior of use cases at various user loads, but also why performance issues arise. For example, if the Login use case degrades at 550 users and exceeds its SLA, is it because of an external dependency, the application server CPU utilization, a database connection pool, a database call, or an application server thread pool? Domain knowledge of your environment and your applications empowers you to be able to correlate metrics and derive accurate conclusions in this section of the Capacity Assessment Report. When I am on-site with customers, I spend a considerable amount of time interviewing them to learn the following: What technologies are they using (for example, servlets, JSP, stateless session beans, entity beans, JMS)? What design patterns have they employed and where? What does a whiteboard sketch of the path of a typical request through the application look like? What objects are cached, and what are those objects used for? What objects are pooled? How is the environment configured (for example, thread pools, the heap, and connection pools)? What is their network topology? What external systems are their applications interacting with and through what communication mechanisms? Through this interview process I cheat : I anticipate where performance problems might occur, so that when I analyze the customer s environment and see them occur in the capacity assessment, I have a strong idea about what metrics to check for relevant correlations. Without this information, constructing an accurate degradation model is difficult at best. Impact Analysis Once a capacity assessment has been performed against a Java EE environment, it should be saved for future comparisons; these performance comparisons are explored in the impact analysis. The impact analysis identifies the differences between two or more capacity assessments with the primary intent of quantifying the impact of code changes against system capacity.
If you are in need for chaep and reliable webhost to host your website, our recommendation is http web server services.

CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING The Use Case Summary columns are defined in Table 9-2. Table 9-2. Use Case Summary Column Definitions Column Description Use Case The use case name or number being presented. SLA Ave The SLA s specific value, or the average maximum value for the defined distribution. SLA Dist The SLA s flexibility value, or the percentage of requests that must fall below the average in order for the use case to uphold its SLA. SLA Max The maximum value permissible for any request, the hard limit (or relative limit if working in standard deviations) that if exceeded immediately causes an SLA violation. Actual Ave The average observed response time for the use case. Actual Dist The observed percentage of requests below the average SLA value. Actual Max The maximum observed response time for the use case. Actual SD The standard deviation of observed response times. Actual 2xSD Two standard deviations of the observed response times. Delta Resp Time Buffer The response time buffer percentage. This is a measure of the buffer that the use case has between the observed average response time and the SLA average response time. It roughly identifies the amount that the use case can grow before it is in danger of violating its SLA. In addition to providing information about the use cases, these sections should also present summary information about pertinent resources. From a Java EE perspective, this information is going to include CPU utilization, heap utilization, garbage collection rates, thread pool utilization, pending requests, connection pools, caches, and request throughput. These four sections present a snapshot of the state of use case response times, resource utilization, and throughput. The conclusion of each section should include an analysis of the raw data, including articles required to substantiate your conclusions. For example, you can include charts and graphs, numerical analysis of the presented data, historical capacity assessment data, and so on. Degradation Model While the capacity analysis sections provide detailed information with snapshots captured at specific points in the assessment, the degradation model reports the entire assessment in a timeline. It identifies trends in response time and resource utilization data throughout the assessment, but its primary focus is on the segment between the expected load and the buckle zone. The degradation model contains a considerable number of graphs, illustrating the following information:
We recommend high quality webhost to host and run your jsp application: christian web host services.

236 CHAPTER 9 PERFORMANCE AND SCALABILITY TESTING What was the frequency of each use case? What was the specific configuration for each use case, such as think-time settings, image downloads, and so on? What was the load testing profile, such as the ramp-up period and pattern, graduated step sizes, load duration, and so forth? How were the physical and logical servers configured in the test environment? What was the monitoring profile? For example, what was monitored and at what level? The test profile serves both as a historical marker for the test context and as a point for later validation against actual production usage. After the application is running in production, vali dating user patterns against the test profile and the observed behavior against expected behavior makes a very good exercise. This exercise can help you refine your future estimates. Capacity Analysis The capacity analysis presents your findings, with all of the details to support your conclusions. This section supplies all supporting evidence in graphs and detailed textual analysis of the graphs. It should begin by presenting the same graph displayed in the executive summary and providing a more detailed overview of your conclusions. The sections following the initial graph detail the behavior of the environment at each critical point in the graph, including sections for the environment behavior at the following times: Expected load The SLA violation point The saturation point The buckle zone Each section should include a table presenting a summary of the behavior of each use case at the expected load. A sample is presented in Table 9-1. Your assessment might introduce the table by explaining, At the expected load of 500 users, the following behavior was observed. Table 9-1. Sample Use Case Summary Use Case SLA SLA SLA Actual Actual Actual Actual Actual Delta Resp Ave Dist Max Ave Dist Max SD 2xSD Time Buffer Login 4 sec. 95% 6 sec. 3.2 sec. 97% 4.7 sec. 1.2 sec. 2.0 sec. 20% Search 3 sec. 95% 5 sec. 2.6 sec. 96% 4.0 sec. 0.6 sec. 1.0 sec. 13.3% Input 7 sec. 95% 10 sec. 5.5 sec. 98% 9 sec. 2.0 sec. 3.0 sec. 21.4% Claim Summary 18.2%
Visit our web design programs services for an affordable and reliable webhost to suit all your needs.