Thanks to the ever increasing density of flash, petabytes of flash storage in a few rack units is now reality. Imagine fitting all of the data center capacity needs in a space smaller than half a rack. The problem, though, is maintaining adequate performance on high density flash system. In this StorageShort we discuss the challenges with maintaining performance in highly dense flash environments.
The speed of the flash itself does not solve most performance concerns. It is the number of workloads that a couple of PB of flash can support. In theory, IT can load all of the data center’s databases a single flash system. The problem is while the capacity of the flash increases, the storage vendor may not have also increased the compute power and networking of the system. And if they did, the system may not be affordable. Traditionally high density flash arrays are ideal for situations where there is a large data set but not a lot of simultaneous workload activity.
The ideal solution is to develop a hybrid system, one that has a high performance access path to low density flash and then a moderate performance access path to high density flash. Active IO is directed to the high performance area, then as it settles it is moved to the high density area.
To learn more about developing a hybrid architecture what our on demand webinar, “Where is the All-Flash Data Center”.