Is it possible for your SOA to become a victim of its own success? If you have built and maintain a fairly healthy stable of services, and other departments across your enterprise begin adopting (or reusing) your services, your servers could quickly become taxed to the max.
In a new post, Dave Linthicum observes that as SOAs grow in scope, so does the challenge of managing the new workloads incurred on existing servers.
As Dave puts it: "As such many SOAs that I'm seeing deployed have a tendency to be out-of-balance or a disproportionate portion of the processing is taking place within a small group of services, and the others are not pulling their load. Thus, you may have the old 80/20 rule again, with 20 percent of the services handling 80 percent of the processing."
Apparently, the folks at IBM have been thinking along the same lines. The other week, Big Blue announced a set of bundled hardware/software products that address SOA overload through virtualization approaches. (Press release here.) IBM says its virtualization technology can shift around workloads in response to spikes in services activities. Hence, part of the growing case for merging virtualization technologies with SOA.
Dave recommends planning for service balancing early in the process, with approaches such as creating a performance model using the profiles of the services, to get a better understanding of the performance trade-offs of your SOA. "Truth-be-told, sometimes you're stuck, and there is little you can do to move processing from one service to another. However, more often than not your services are 'balanceable' and you should consider how the processes are distributed across services, as well as logical organization of those services."
Leave a comment