Two years ago, we completed an ambitious test of nine APM vendors over the course of almost a year. We designed a thesis around holistic APM products that not only identify problems but take corrective action to resolve performance issues before they impact users and customers. Actions may include allocating additional bandwidth or server processing capacity, even rolling back configuration changes.
Where corrective-action capabilities are lacking, organizations will struggle with fixing problems manually, though fast identification and notification are better than nothing. We also suggested that for truly holistic APM, products should locate and predict performance problems across all application components and support all facets of the app infrastructure. Service modeling was another key area; we sought to model critical application services and report on compliance with service-level agreements.
Over the course of the last two years, the product vendors have gone through mergers, acquisitions and made software enhancements to improve their systems. However, the most radical change to the APM market has been the explosion of cloud applications. Cloud computing, with its resource pooling architecture really turned the APM market on its head. As organizations continue to develop use cases and pilot cloud environments, service level agreements around applications will be a tough nut to crack. All of the issues we addressed around the challenges with holistic APM become even more complex in a cloud environment. And with the reality of the hybrid cloud, combining both internal and external resources, having a clear view of the infrastructure is not possible with the cloud architecture of today. Few enterprises are willing to trust the cloud vendors completely.
So, what do we do? As IT managers we need a new paradigm for APM. There are plenty of products that do a good job of monitoring the performance of apps in the cloud; the issue is that they really can't do anything if they find a problem. They also can't determine what really the root cause of the problem is. On top of those shortcomings, the murky SLAs from the cloud vendors are really designed to protect themselves, not their customers. So, even if you find a problem - getting a credit may take some effort.
Instead of looking at monitoring and measuring every aspect of the applications, we need to turn our attention to new architectures that will utilize the cloud in different ways. Cloud bursting may allow the rapid expansion of resources to meet demand preventing performance issues. WAN optimization can allow better access to the cloud apps using caching and compression over the network. These are just the start of new ways to use the cloud to achieve better app performance. While we are still in the early stages of cloud apps and managing performance in that environment, we need to start changing our psychology around management and measurement. Ultimately, we are after better app performance, not building a complex IT infrastructure to manage the app environment. APM is dead! Long live APM!