Tuning WebLogic with Third-party tool tools

As the web becomes more and more demanding on the performance of the system, the key to any system is how to ensure the performance of the system and can predict and locate the problem before performance problems occur. The system test before the system is on line and the performance monitoring of the whole system is the method to ensure the system to run with excellent performance.

At the end of last year, I wrote an article about how to use Jprobe to discover and locate the performance bottleneck in the application of Java EE, Jprobe is a Quest performance optimization tool for the application in the development process, but this does not meet the overall performance monitoring and management requirements of the system mentioned above. In response to this requirement, combined with current market performance analysis, tuning and management tools such as IBM Tivoli, HP OpenView, the main function of such tools is to manage the entire system; others, such as Wily,veritas i3, Such tools also have the ability to manage and monitor the whole system, and have excellent tuning and monitoring capability at a certain technical level; Other tools such as Quest Jprobe, as described above, are primarily for program-level performance optimization during development.

This article will combine wily and weblogic to describe how to use the wily tool to monitor and manage a distributed system in a comprehensive way with the current popular application architecture. In the past, many of the tuning of the Java EE are based on developers or manufacturers of technical personnel according to experience to the problem positioning and tuning, can not do a full range of understanding of the system. With the help of wily, we can proceed from the customer experience to a specific SQL statement for in-depth analysis, to complete the system performance monitoring and management.

Wily company was founded in 1998, its first investment side is BEA, has the very good support to WebLogic.

The core product of Wily is Interscope, including Introscopeenterprise Manager, Introscope Agent, Introscopeworkstation. Through Introscope can clearly show what happened to the Java EE application, such as what the problem, for example, when the applicability of the decline, to identify the location of the Java application system to cause problems is a very cumbersome work, with the help of Introscope will change Very simple.

The wily Introscope system architecture is shown below

Characteristics of Wily Introscope

1. Application monitoring, low system overhead, overall monitoring system performance, proprietary blame technology to instantly and accurately identify the components that cause performance problems;
2. Application server monitoring, the most extensive application server support, Autoprobe technology Integration Application Server automatically obtain server information; multi-platform support;
3. For non-Java system monitoring, through the EPA and the Data API module, 100% thread safety;
4. System Management process integration, through the MIB and System Management Framework integration, the console can be integrated with HP Openview,tivoli, Smarttrigger alarm system and e-mail, procedures and other integration;
5. Wily unique Smartstor technology can facilitate the fast storage of monitoring data information, and can complete the system historical data analysis and trend analysis according to the data information;
6. Some proprietary technologies have become the industry's standard.

Through the Introscope structure diagram, we can see that the core is Introscope Enterprise Manager, through the various agents deployed in the application to collect the performance data of the system running, summarized to EM for analysis, and can use the analysis of historical data to evaluate the performance of the system in the future; the analysis of the structure can be specific to what position in addition to what the problem, and the problem of classification feedback to the corresponding system maintenance personnel, such as the network, the system hardware maintenance personnel, or development and testing personnel, to the problems arising from the adjustment.

Integration of Wily and WebLogic

Wily has a WebLogic performance monitoring template for Powerpack, effectively monitoring the most critical weblogic resources, including the thread pool, JDBC connection pool, and the first to implement the PORTAL (BEA PORTAL,IBM Portal, etc.) performance management and monitoring. By Powerpack you can see the various performance metrics for the applications deployed on WebLogic, with WebLogic medical records example, as shown in the following illustration:

You can see a variety of performance indicators, including system resources, and Java application of various components of the performance indicators, through configuration can be traced to a specific JSP or servlet performance, and can be configured to a certain performance indicators to achieve the specified threshold after the alarm operation.

The transaction trace function is provided to analyze the internal conditions of a specific transaction over a specified time.

The tree structure can be seen within the transaction and quickly navigate to a problem of the operation, through this technology can be real-time tracking a specific transaction problems in the production system, providing the transaction execution path and component response time details, such as the above figure. And can fix the performance problem of the transaction in time.

In addition, the Powerpack package provides a Performance view console for the WebLogic system, which has problems with the intuitive monitoring system, and allows easy customization of the various performance metrics that are concerned with the console. After customization, you can view the operation of the whole system through the browser.

Configure startup steps

1. The installation of BEA WebLogic (Install the example program with the installation belt, the detailed steps slightly);
2. Introscope installation, after the installation of Introscope Enterprise Manager,introscope workstation,introscope webviewer three components (installation steps slightly);
3. Powerpack for WebLogic installation.

Installation steps, simply unzip the Powerpack package to Bea's installation directory (other directories can also be specified at the time of configuration);

1. Configure startup WebLogic
   As an example of the medical records service with WebLogic, the configuration method is as follows
   Modify boot file Startmedrecserver.cmd (different JVM configurations are slightly different)
   

   Set java_options =-xbootclasspath/p:c:\bea\weblogic81\wily\connectors\autoprobeconnector.jar;   C:\bea\weblogic81\wily\Agent.jar   -dcom.wily.introscope.agentprofile=   c:\bea\weblogic81\wily\ Introscopeagent.profile

2. Start Introscope Enterprise Manager;
3. Start WebLogic;
4. After service starts, after landing through Introscope workstation, you can see the monitored application nodes and related performance nodes in the console, and then you can see the corresponding performance index value after the expansion.

After the configuration startup is complete, the system can be viewed in a variety of different types of charts in the console by setting the appropriate monitoring performance entries.

How do I find system performance problems?

As the example program in operation does not have great pressure, through the console can see the load of the system, in order to simulate the production system and the same operation, Here, with the help of Segue Company's Silkperformer to the WebLogic example program to pressure (Silkperformer can also monitor the performance of the system indicators, where the collected data and wily for a ballpark comparison), Analyze the performance problems that may exist in the system through the performance data collected under different loads.

Simulate run steps

1. Run the WebLogic service with good configuration;
2. Select the business operations that you want to simulate, such as selecting a login operation.

　　

1. Recording this business operation with Silkperformer, generating the script of simulating pressure;
2. The pressure mode settings, such as the figure, set different number of concurrent users.

number of concurrent users	Test Time
10	10 minutes.
20	10 minutes.

Pressure mode setting

Possible performance issues and observed performance metrics feature descriptions

Performance Issues	Performance Index features	Problem Description
System performance decreases with increasing load	The number of threads configured by WebLogic is scarce with increased load.	Resource bottlenecks
Predictable deadlock, the system performance decreases with the increase of running time gradually	Resources such as JDBC connections cannot be recycled, as you can see from the performance metrics diagram that the resource is 0 and there is a lot of waiting.	Resource leaks, etc.

10 Concurrent system thread usage, obtained by wily

Pendingrequestcurrentcount=0,waitingconnectioncurrentcount=0, indicating no waiting request, the system responds quickly.

JDBC usage of 10 concurrent systems, acquired via wily

The maximum value of the Concurrent invocations is 8, and the time curve of the average query is also fairly stable.

20 Concurrent system thread usage, obtained by wily

Executethreadcurrentidlecount=0,pendingrequestcurrentcount began to change, comparing the threading usage of 10 concurrent users, it was clear that the system's thread resources began to be low under 20 concurrent pressures.

JDBC usage of 20 concurrent systems, acquired via wily

Average query time has a relatively large fluctuation, after running for a period of time you can see that the value of 0,connection count is also kept unchanged, basically do not respond to get connection requests. This time access to the system page, cannot enter.

Conclusion

The above example only through simple performance indicators to observe the operation of the system, for a complex system also needs more performance metrics data to analyze whether the system is running well, such as can check whether the system has memory leaks, network speed is fast enough. The general system tuning many are in the problem, with experience compared to the performance of the system to carry out, many times may spend a lot of time to locate the real performance bottlenecks, with tools, can be intuitive to the entire system of various parts of the monitoring, once there are problems, Can timely alarm and can quickly locate problems to solve problems.