Application Management is the most complex discipline of technology management.   An application depends upon every layer of the OSI stack functioning correctly, the internal application working, and the end-user operating the application correctly.   Relying on all three of these elements working concurrently indicates that application availability is likely a common problem for most organizations.

Visibility is typically the biggest problem; most tool sets cannot handle examining all three.  You then lose some perspective, make erroneous assumptions and increase downtime. The first area that should be address is application dependencies – the network, servers, and database.   These components do fail, and the application is dependent upon them and thus they need to be monitored at least from an availability point of view.   Most outages are caused by this area, and it’s best to track this so that application resources are not assigned to fix a problem that is outside of their control.   The inverse is also appropriate; when everything is available you can concentrate on applying application development resources.

The second area is the most overlooked – the internal application.  Typically this is usually a multitude of log files that consume countless man hours to look through.   Some mission-critical applications, however, may have a custom written tool set to show/store custom tracking data, and some can forward that data to a centralized management solution.   This data is absolutely critical for quick resolution to application problems.  With this data, it is sometimes possible to migrate some resolution capabilities from the application developers to operation engineers, making resolutions even quicker.

The third area is becoming more popular, but provides the most misleading data – end-user experience.   Monitoring the end user is typically done in one of two ways: synthetic transaction or passive listening to application transactions.    Tracking failure rates and errors is key to better application management, but performance response times also are indicative of problems.   However, what is misleading in these metrics is that they are opposing sides:  synthetic transactions are too sporadic, and focus on a subset of interactions; whereas the passive monitoring sees everything, even confusing actions which can cause distorted response data.   This area is very important for quality control, but false positives are a common occurrence here.

Holistic application management is complex and difficult but necessary to maintain mission-critical applications.   If your business is depending upon the health, accuracy, and performance of an application (ASP/eCommerce/etc) it’s the only way to go.

If you want to learn more about application monitoring and how Monolith takes a dramatically different approach to it, I encourage you to read our application monitoring datasheet or visit our web page.

Technorati Tags:
application management, application monitoring, end user experience, application stack, technology management, networking monitoring, network management, synthetic transaction, passive monitoring, BSM, SLM