There are many ways to approach Application Performance Management (APM), depending upon who is looking for the diagnostic capability and what the exact objective is. There are three different APM approaches in all:

• Network-based APM
• Agent-based APM
• End User Experience Monitoring

These methods each have different advantages and drawbacks.

Network-based Application Performance Monitoring

The network-based application performance monitoring approach consists of capturing network traffic and analyzing application exchanges to provide statistics on network performance, such as server response time. From this analysis, performed from OSI layers 4 through 7, we can:

• measure the end user experience
• understand how applications are delivered through the network to all users by capturing traffic in the datacenters
• provide a historical and real-time view of performance
• determine the origin of a slowdown (between server, network and data transfer)
• analyzing a vast number of applications at a time with no upfront configuration
• provide a transaction level performance metrics with no performance impact

Network-based APM is great way to provide a wide-angle view of the performance of all applications traversing the network with an easy and passive integration in the IT infrastructure. (SkyLIGHT PVX is an example of a network-based APM solution.)

Code-level Application Performance Monitoring

This approach consists of deploying software agents on the servers of the key application chains. These agents will collect response time and error information on the transactions processed by the servers.

They are great for:

• identifying which part of the application code is faulty or slow
• allowing administrators to drill down to the code level

They are mostly used in development, pre-production, and Q/A phases.

End User Experience Monitoring

This approach consists of deploying software robots that playback predefined scenarios (which are representative of common user actions) and measure the overall execution time for a scenario. This works for a set of protocols and applications.

This is great for:

• providing reports on execution times that match perfectly the vision of performance of end users.
• measuring applications with a limited variety of uses (always the same scenario)
• measuring a limited number of business-critical applications
• testing performance from different locations

This is normally deployed for the most critical application(s).

Feature comparison