Managing the performance of apps in the cloud calls for special measures.
Applications in the cloud? Now we're talking. Talking complexity that is, of a degree that far outstrips what we had to deal with in the (possibly) good old days of apps running on monolithic hardware platforms. In a computing cloud, no matter whether it's private or public, there are more moving parts, more changing relationships between those parts, and more metrics to collect, organize, and interpret as quickly as possible.
Hence the need for a new generation of application performance management (APM) tools specially attuned to the workings of cloud-hosted apps.
As Julie Craig of Enterprise Management Associates (EMA) notes, troubeshooting of apps in the cloud is much more of a challenge than it ought to be. One reason: Providers of cloud computing services don't always share the kinds of performance information that's needed if IT managers down on the ground are to keep apps running full bore and keep users happy.
To help make the troubleshooting problem clear, Craig dissects the many layers and potential bottlenecks in a cloud-hosted software service. Any number of factors could affect that service: congestion on the wide-area network, contention in the user's client computer, sluggish message-passing between different IT services making up the app, failed load-balancing, database issues, and more.
In short, who're you gonna call when SaaS slows down? With so many layers of function contributing to the application's performance, it's only with a thorough metering of full infrastructure that IT has any hope of zeroing in on problems when they occur.
Craig distinguishes among several different approaches to APM, each with its own strengths and weaknesses. There are, for instance, what she calls top-down solutions, such as running synthetic transactions. This approach provides a good summary measurement of how a particular app is performing at one particular moment as perceived by end users, but it doesn't help diagnose any slow-downs. The bottom-up approach, meanwhile, provides visibility down into the guts of the elements comprising an app, but it offers none of the "end-to-end perspective" that is vital to managing a mission-critical app in the cloud.
Ideally, an APM tool will be able to plug directly into leading providers' cloud services, whether those are delivering infrastructure, platform, or software services. This will enable the tools to collect highly useful metrics about the service provider's operations and not just treat the provider as a "black box" whose inner workings are not accessible to view.
Right now, to one degree or another, all providers of APM tools are scrambling to "cloudify" their products and enable them to provide end-to-end, holistic views of cloud-hosted apps and their performance. The winners, it appears, will be those that can deliver the right mix of function, comprehensive architecture, and cost.