Primary storage

The storage in which production workloads live. Examples include live virtual machine workloads and IO-intensive workloads – both of which are used often and need to be accessed quickly.

Example storage technologies, Dell Compellent, Dell All-Flash ME4 storage

Secondary storage

Storage in which infrequently accessed objects are held. For example, file server data that is older than a certain age, low priority virtual machines and workloads that do not require many resources.

Example storage, Dell spindle based ME4

Archive storage

Storage that does not need to be accessed regularly. For example, files that are not in use, but may be required for compliance reasons, and secondary copies of backups.

 Example storage types, Dell Data Domain storage

Cloud Native storage

Storage that conforms to the object storage policies dictated by the public cloud hyper scalers. Typically referred to as S3 compliant storage. This can be used by applications that require object storage. A typical use case is as a secondary target for backup data.

Example storage type, Dell ECS storage

Sizing of the storage depends on the engagement and what it’s for.

Typically, it’s for a VMware infrastructure refresh or backup target storage.

The key thing we need to identify is the characteristics of the workloads that will sit on disk.

By this I mean, we need to know the block sizes of the workloads, how much data there is, the average change rate of data and the read/write profile of the data.

To do this, we can use a couple of tools. These include Dell Live Optics which captures that information from a server environment, or, we can use Veeam ONE to monitor change rates for backup sizing.

In particular, if it is for backup storage, we need to know how many copies of the data we need to retain and for how long. Recovery Time Objective will come into play here, as well. If RTO is quite long, then we could look at some slower or archive storage to host backups.

The main problem people come to us with are complaints of slow performance. Following the exercises detailed above, we can pinpoint where any bottlenecks are in the environment. It is not always a storage constraint; it might be CPU or memory, or just misconfiguration of something.

This is where we can add value. We don’t throw something new and shiny at the problem in the hopes that it goes away – we study and analyse the problem and come up with the correct resolution.