Feature

Digg This!

Enterprises rely on IBM's WebSphere Application Server to quickly build complex applications and highly scalable transaction services. As such, the business criticality of these J2EE applications has grown. WebSphere's importance will continue to expand as businesses look to Web services as the next evolution of application architectures.

Trusting your business to this future means that IT must guarantee a quality end-user experience by managing the application server and application components' performance. This challenge is daunting because proactive administration of complex and dynamic component interactions is required. It can be overcome because a solution is now available to help IT proactively manage the availability and performance of WebSphere Application Server and deployed applications. This article looks at the challenges of managing these complex environments and illustrates some of the solution features that give IT the power to ensure the ongoing success of WebSphere Application Server as the heart of the J2EE application architecture.

Managing WebSphere-based Applications Is Business-Critical
WebSphere Application Server (WAS) has become the foundation that many enterprises rely on to connect to their customers, partners, and suppliers. It's used for applications as diverse as the number of companies using it; however, in most cases businesses use these applications to drive their profitability. For example, one tire manufacturer off-loads its 30,000 phone orders per day from dealers to a WebSphere-based solution to reduce costs; ensuring high performance is critical to achieving this goal. The flip side of these success stories is that it isn't possible to ensure high performance if the application's execution is invisible. This is the case for most IT organizations implementing WebSphere-based applications. The few embedded tools to monitor and control these critical applications have been slow to arrive and a problem has arisen: if IT cannot ensure high-performance applications, the business dependent upon them cannot thrive.

Managing WebSphere-based Applications Is Challenging
The difficulties arise because WebSphere-based applications are both critical for business success and among the most difficult technologies to manage. There are three main reasons for this:

• Java application architectures are still rapidly evolving.
• Current management solutions often fall short.
• Custom EJB and servlet components are developed under strict time-to-market pressures and are not instrumented for management.

First, there is the rapid maturation of the technology from a servlet-based design to full-blown J2EE component architecture. The rise of interest in WAS lies in its support for other object technologies (CORBA and COM wrapping), transaction services that work with CICS, integration with MQSeries and adapters for SQL databases, ERP, SCM, and CRM packages. Developers can rapidly create EJBs and servlets to deliver new business services. Inherent in such component architectures are software dependencies that affect service performance. These dependencies aren't easily visible because traditional application management analytic processes do not apply, making it difficult to achieve the consistent application performance that leads to maximum business value.

Secondly, IBM has tools that provide visibility into the WAS infrastructure. WAS Resource Analyzer monitors the execution of EJBs, servlet resources, and Java Virtual Machine (JVM) resource usage. Log Analyzer allows administrators to sort WAS log entries based on severity, process ID, thread ID, and so on. Site Analyzer provides basic Web-site traffic measurements. The value of these tools is that they provide detailed information about the WAS infrastructure that administrators can use for performance analysis tasks; however, they don't:

• Automate and therefore speed the problem identification and resolution process: With these tools administrators still need to create a test case to reproduce the problem and then analyze execution and log information from every component to isolate the problem.
• Provide visibility into the overall infrastructure (e.g., back-end systems and databases) and the dependencies among the infrastructure components: Only by managing in this broader context can the application server's health and performance be correlated with the health and performance of other components and with end user-experience data so that IT can truly understand the magnitude of a problem and its root cause.

The inability to automate and provide end-to-end visibility results in longer downtimes or performance slowdowns that negatively impact business. Complicating matters is that many third-party solution providers focus on extending management functions to the WebSphere Application Server software itself. They don't concentrate on the custom EJB application components that encapsulate the business logic. As such they have difficulty providing visibility into the application components that deliver the actual business value. No visibility means no possibility of improving the performance of the most critical parts of the application.

Finally, companies choose to build custom applications rather than buy off-the-shelf software only when the application features are critical to delivering a competitive advantage. In other words, only the most important business functions are built internally and usually under extreme time-to-market pressures. Companies speed their development processes by leveraging the noncore functions supplied by WAS and focusing all of their creative efforts on the critical EJB components. This need for development speed, however, means that there is no time for rigorous quality assurance, performance testing, documentation creation, and technical support training. In many enterprises there is inadequate interaction between operations and development teams, which exacerbates the problem.

This compartmentalization of knowledge means that all too often no one truly understands how the custom components interact with the underlying WAS in production environments. The result is that operations inherit applications without code-level knowledge to troubleshoot runtime problems. Without a feedback loop, developers don't know when and which components fail to meet business performance requirements. Thus, optimizing components' execution for real-world conditions cannot be done effectively without tools that facilitate developer-operations communication in building and tuning high performance application components.

Enterprises that can handle these difficulties will avoid the revenue losses, user satisfaction issues, and spiraling cost of inefficient management associated with under-performing J2EE applications. Enterprise IT needs a solution that delivers component-level visibility into and control of IBM's software, the business-specific EJBs and servlet components built on the WebSphere platform. The solution must also correlate the application's performance within the broader context of the entire infrastructure delivering the service and its relation to the users' experience.

Requirements for a Successful Management Solution

To address these problems IT requires a management solution that:

• Monitors end-user experience metrics against the business's goals
• Delivers detailed knowledge of the behaviors of specific technologies
• Captures the dependencies of the multiple technologies that deliver a business application
• Correlates the understanding gained through these three steps to rapidly pinpoint the root cause of problems
• Enables advanced manual or automated problem resolution actions

One example of a management solution that fits these requirements comes from Resonate. Resonate Commander Solutions relies on a service model to drive proactive management of the business services delivered to end users. This service model captures the business objectives and the relationships between the various technologies delivering those objectives. When Resonate Commander Solutions monitors the relevant technologies and end-user metrics it correlates the data with the service model to determine and correct the issue. For example, when diagnosing a performance problem with a catalog service, it uses the service model to automatically determine that the associated server operating system, relevant database, and network routers are operational, but the WAS application is causing problems. This service diagnosis eliminates the finger-pointing that traditionally lengthens the troubleshooting process and adds to downtime costs.

When selecting a management solution, consider a solution that continues its analysis by drilling down to identify the offending servlet and then notifies the application manager of recommended actions. Usually, management solutions accomplish this by providing an application-specific module that works within the broader solution to provide in-depth, service-based management capabilities to WAS administrators. This module would:

• Provide deep knowledge of WebSphere Application Server and custom component performance through its support of both IBM's proprietary APIs for WAS version 3.5 (e.g. Enterprise Performance Management and WebSphere Control Program) and their open PMI interface for WAS version 4.0. Through these, the module would access the same detailed information delivered by IBM's native monitoring tools (e.g., Resource Analyzer) to troubleshoot problems within the larger context of the actual WAS deployment.

• Apply the service model capabilities to capture the relationships between the various components that make up a WAS instance such as stateless EJBs, JMS sessions, JDBCs, etc. WAS administrators can, therefore, easily and quickly drill down to identify not only the specific offending EJB, for example, but also other things that may have been impacted, and perform actions on specific WAS instances or their dependent components.

The management solution, therefore, must combine both specific component details and the broader service context to identify problems and automate resolution of the specific WAS instances. This WAS-specific management data must be automatically rolled up into an end-to-end service view allowing administrators to effectively manage the WAS components in the broader context of end-user experience.

A more sophisticated management solution also gives IT real-time control over the production environment through a variety of corrective actions. IT managers should be able to specify the type and sequence of actions, such as restarting failed application processes or redirecting transactions to an alternate server until the problem is resolved. With these capabilities IT no longer simply reacts to service problems - it prevents them from occurring.

Business Value of Management Solutions
Business performance now moves in lockstep with WebSphere application performance. IT, therefore, needs solutions able to both manage the user's quality of experience and troubleshoot specific technologies. A management solution that can aggregate data across an infrastructure stack (components, middleware, back-end systems), and across the IBM "stack" (WebSphere Application Server, MQ Series, and DB2) into a single service model, are invaluable because IT managers can rapidly pinpoint the infrastructure components negatively impacting the quality of experience. A solution that can then drill down to provide granular visibility into component performance enables WebSphere administrators to rapidly and accurately troubleshoot technology-specific root causes. Taken together, these abilities minimize the financial impact of technology problems on the business. This translates into lowered costs, revenue saved, faster business transactions, and improved customer satisfaction.

The ability to automate problem resolution for common problems gives IT proactive control over the capacity and performance of the underlying infrastructure. This improves the efficiency of IT staff, allowing reallocation of limited resources to strategic efforts. The automation also allows business managers to maximize use of available resources providing tangible control over the quality of service experienced by external users.

A more sophisticated management solution would also facilitate the "holy grail" of application development - code reuse. Component architectures have long promised reuse as a business benefit, which few enterprises have realized because the impact of specific components on overall application perfor-mance has not been visible. Developers building these high-performance components could not be rewarded and given the key to reuse the components. Management solutions give operations, architects, and developers a common language to discuss performance and development issues. This feedback loop delivers better understanding of component behaviors, faster tuning of under-performing components, and identification of best practices for component building. These benefits give businesses a means to make component reuse more of a core endeavor for their development staffs. Also, a solution that collects granular information facilitates long-term planning, general system tuning, and development. By knowing, for example, the amount of time an EJB spends in a queue, the most failing EJB, or the most invoked EJB, IT can optimize both the production infrastructure and application code to prevent failure altogether.

Furthermore, with the rising importance of Web services as executable components on the application server platform, enterprises must now also incorporate Web services as a well-understood and managed part of their application architecture. A solution that provides granularity of visibility and control provides the means to measure and manage the quality of service of Web services. Web services deployed on the application server will be visible, measurable, and manageable so that IT can confidently deliver high-quality Web services wherever they are needed.

Conclusion
Enterprises cannot fully realize the business benefits delivered by their IT infrastructures without proactive management capabilities. The resulting business risks are enormous, as WebSphere-based services determine the business' ability to compete. Three degrees of IT manageability are necessary to control these risks:

1. Proactive management of the business services: Monitoring, troubleshooting, and correcting infrastructure problems based on their impact and role during the delivery of the business service
2. WebSphere Application Server management: Detailed visibility and control not just of the application server itself, but the custom components that encapsulate the business ser-vice.
3. Correlation with end-user experience: Enabling management from the end user's perspective.

IT needs a management solution that provides the power to make all three requirements a reality for the enterprise.