I just finished reading this post from Heidi Biggar (HDS) about importance of storage in virtual environments.
The post starts with an impressive graph showing that near 80% of enterprises are migrating or planning to migrate their Tier 1 applications to a virtualized infrastructure! Great, but to realize a full virtualized environment you need a fully virtualized infrastructure: Storage needs to be virtualized too!
To simplify I can say that virtualization of storage can be done, at least, in a couple of ways:
- Hardware side: means the ability to virtualize storage arrays (like IBM SVC, HDS USP or NetApp V series) and present them to the fabric transparently adding features from the virtualizer.
- Software side: enabling software features like wide striping, thin provisioning, space reclamation, transparent migrations, and so on.
I would like to add an important feature, very useful in virtualized environments, and not listed in the Heidi’s post: Automated Tiered Storage.
why is it so important?
With VMware you will have VMs and related data disks stuffed in datastores (datastores == vmware volumes and in a coarse way 1LUN == 1 Datastore), so you will have volumes with tens or hundreds of VMs and related data: it will be very difficult to avoid hotspots without Data Progression or a lot of manual work!
Moreover, Moving Tier 1 applications to virtual means that you have already migrated Tier 2 and less valuable applications before! Probably, you will not create a second infrastructure, you wiil rather consolidate the existing one. When your virtual infrastructure will be full populated of T1, T2 and T3 VMs the problem described above will become very huge.
Last but not least, SSDs! Yes, because when you will add T1 applications to your virtual infrastructure, performance will be one of the most important issues. With a feature like Data Progression you can buy few SSDs (now very costly) and obtain speed without thinking about layout of LUNs and space constraints.