A composable storage system can be virtually subdivided into independent arrays that provide more specific control over the allocation of performance and capacity. As opposed to quality of service (QoS), which tries to make sure certain applications get the performance they need from the storage system, a composable system assigns specific controllers and drives to an application or an environment. It provides a more reliable and consistent guarantee of performance than does QoS.
As we discuss in this StorageShort, the capabilities of composable storage are particularly interesting to organizations that need to insure service levels across a wide range of applications or operating environments.
Service providers are at the top of the list for the discrete assignment of performance that composable storage offers. They can invest into a single scale-out storage architecture, then sub-divide its compute and storage resources to customers based on subscription level. Those customers could then leverage QoS within their allocation to make sure that certain applications get the priority performance.
No Noisy Neighbors
A common concern among service providers is when one customer has an application that consumes all of the available resources. Known as the “noisy neighbor” problem, providers are often forced to over allocate storage compute and storage media in order to combat the problem. While QoS does help the noisy neighbor problem, it is really only managing available resources and does not typically manage a situation where two high priority workloads both spike at the same time.
Composable storage solves this problem by hard allocating storage compute and media to specific customers. Their use of resources can’t impact other customers since they don’t have access to the other customers resources.
To learn more about composable storage, watch our latest webinar, “NVMe, NVMe over Fabrics and Beyond – Everything You Need to Know”, where we discuss how NVMe technologies will power composable storage systems.