The problem of resource contention in the IT world is not new. Contention is what happens when demand for a shared resource exceeds the supply. Basically, when several requests are racing to utilize the same resource, one of them will win and the rest will have to wait. Without considering contention in a given service or situation, performance degradation will result. In a world where costs are always a factor, addressing resource contention with regard to CPU, memory or storage is very important.
It’s also important in the cloud. When your company is operating a local data server, managing the problem of contention is entirely in your hands. You can evaluate whether the performance issues you’re seeing are actually resource related. Then identify which transactions are competing for resources and prioritize appropriately based on business needs.
In the cloud however, where you are paying someone else for their online services, often the most you can do is complain about performance and trust they will resolve the issue to your satisfaction. Contention and what is called the “noisy neighbor” issue is one of the reasons companies often give for staying out of the cloud. The argument is that another company sharing the same cloud provider might create a workload that could put high demands on the server or database and impair your performance.
However these problems, when they occur, are rarely the result of simply being in the cloud. Often they are actually the result of a poor cloud provider. If you think about it, cloud services providers; Saas, IaaS or PaaS are in the business of reliable, dependable service and work carefully to isolate workloads from each other. Though contention is an issue for any shared system, local or cloud, the problem is usually not a cloud issue, but a provider issue. In this regard, carefully choosing a provider is essential.
It’s important when evaluating a provider and their service to run workload benchmark testing using the type of implementation you are planning. The results of those tests on the actual cloud platform should be within acceptable tolerances. If they aren’t, you should keep looking for a service that gets the job done.
Contention matters, but don’t blame the cloud for impaired performance. This has the potential to happen anywhere that there are shared resources. It’s up to you to manage the problem, whether on your own local turf or on a cloud service platform.
