Understanding Performance Problems

Feature Article | June 13, 2007 by admin

Does this sound familiar? Users are reporting erratic response times. Their transactions are processed at different speeds for unfathomable reasons.
Operators of networked Web applications are all too familiar with this problem. Such problems require immediate attention, but that’s easy to say, hard to do. Finding the cause of the problem is like finding the proverbial needle in a haystack. After all, the cause may be the browser, SAP NetWeaver Portal, the Web application server, a back-end application, or the network.

Monitoring from the application layer

Examining all the components individually would be extremely inefficient. And up-front load tests are just as unhelpful because you can’t entirely simulate real-world conditions. The best solution is centralized enterprise application management (EAM) software that continually monitors all the components of networked Web applications to identify the causes of problems. The need is especially acute if you need to monitor the performance and availability of complex, business-critical Web applications based on Java. With complex, Java-based applications, processing usually involves a number of different server and back-end applications.
EAM differs from traditional system management, which manages individual components of an IT infrastructure, such as the network, servers, or databases. EAM concentrates on the quality of transactions from the viewpoint of the application layer. It answers questions like the following. Which components accept a transaction request? Which paths within the Java Virtual Machine (JVM) – the platform of the Java components – are used? Which back-end systems are addressed from the JVM and with which transactions and response times?
The software must determine which resources were used and when, how often, and why they were used. From the viewpoint of the application, the interplay in this context between individual or several JVMs is also important, as is the behavior or the back-end systems.

Time is money

The implementation of enterprise performance management quickly pays for itself in companies that operate complex, business-critical applications. It helps them improve the stability of their applications, minimize the effects of problems with application performance, and lower overall operating costs. When you can also transform your processes from purely reactive performance management to proactive performance management, you can avoid critical failures or performance issues in your production system.
Recent studies give a convincing answer to the question of price. Analysts at the Butler Group estimate the direct costs of poor performance of limited availability of software in industrialized nations at least €70.8 billion. The German economy alone loses €4.7 billion each year because of poorly functioning and monitored IT applications.

A look at the inner life of Web applications

EAM across all applications

EAM across all applications

CA Wily Introscope for SAP NetWeaver provides users of SAP software with an EAM solution that is also used by the SAP Active Global Support organization. According to experts like Dr. Uwe Hommel, executive vice president of SAP Active Global Support, the software provides the most advanced functionality for Java diagnostics. The functionality includes the availability and performance of Enterprise Java Beans (EJBs), Java Server Pages (JSPs), classes, and so on along with measurements on memory and CPU usage. The overhead of the tool itself is hardly an issue. Thanks to patented Blame technology, bottlenecks can be isolated in Web-related applications right up to the level of individual Java components, including servlets, JSPs, EJBs, JCOs, classes, and methods.
The name says it all. It’s a reference to the “blame game” played by support staff in an organization. Each player blames the other’s application components as the cause of the problem. In a sense, blame technology opens the inner life of an application with byte-code instrumentation (BCI). The JVM runtime environment does not require manipulation, which differs from the logging approach – the log entry in the source code. BCI sits directly on the byte-code level so it can be executed transparently and correctly by every JVM that is faithful to specifications.
An elegant implementation of BCI captures informative performance metrics up to the method level of the Java classes with minimal overhead and without destabilizing execution. If needed, the metrics can be loaded on the fly and queried at regular intervals. The data of the metrics can be correlated and presented as the result of the application environment.
The ability to use monitoring at the byte-code level to view events in running Java applications provides quick information on the “guilty” components. Within a few hours, you can identify errors in the operation of a business-critical application, as well as errors caused by incorrect settings in memory parameters of the Web application server or by a faulty database index on the central server.

Integration with SAP Solution Manager

Monitoring Web transactions

Monitoring Web transactions

SAP provides its customers with CA Wily Introscope with SAP Solution Manager. SAP Solution Manager is the application management solution offered by SAP for implementation, operations, and monitoring. It includes functions for the configuration of SAP Business Suite applications and operating-support functions for change requests, test and incident management, root-cause analysis, and solution monitoring.
SAP has integrated CA Wily Introscope into its own standards for root-cause analyses and solution monitoring. For example, activation of end-to-end trace, a function of SAP Solution Manager that takes detailed measurements of the technical activities of an individual user, activates the transaction trade of CA Wily Introscope. That allows you to determine where a performance problem is hiding in a technical component of a scenario. End-to-end workload analysis in SAP Solution Manager records statistics of traditional enterprise resource planning (ERP) solutions and the performance data of Java and Microsoft .NET-based components to provide one view of the performance of various systems in a customer’s solution.
CA Wily Introscope is built on an agent-manager architecture that supplements the workplace and tool environment for administrators. SAP Solution Manager contains diagnostics for SAP NetWeaver that enable rapid implementation of the required techniques. A variety of preconfigured CA Wily Introscope agents capture measurements on the performance of components within the Java and J2EE application and the application server of SAP NetWeaver to be monitored. The agents report the measurement data to the Enterprise Manager of CA Wily Introscope, the central collection and control component installed on the diagnostics server. SmartStor, a kind of flight recorder, helps the Enterprise Manager of CA Wily Introscope record the data continuously. Historical data can also be written to a simple text file or stored in a relational database.
Data can be displayed in a browser interface, WebView, at any time in personalized dashboards or in concise reports on problem solutions, trend analysis, and capacity planning. The CA Wily Introscope workstation also allows to define specific alerts for individual measurement data or groups of measurement data. Users can adjust views as necessary. CA Wily Introscope also contains ready-to-use reports on application performance, capacity planning, and the management of service-level agreements.

Erich Lauterbach

Tags: , ,

Leave a Reply