Storage Futures Weds 16:30

Today's problems:
- how do we add more capacity without down time?
- how do we manage fault isolation?
- planning and sizing over time is difficult, paticularly with VMs which are dynamic in nature
- how is capacity optimised?
- is the storage SLA matching the application requirements?

Scale Out
- add more arrays to virtualized storage means trivial hardware refresh, scales with servers, difficult to produce differentiated services when all the VM hosts can see all the same datastores
- example is web apps being on one cluster that has no cross site storage replication and those apps that need hot / warm are in a different cluster that has cross site storage replication


VMware Vision
- VM profile of storage specification allows VMs to have different policies within the same cluster
- Define back up and snapshot policy
- The policy is shared with the storage infrastructure so the storage knows what to do with the VM


Desktop Vision
- VDI cheaper storage than desktop
- storage simpler to manage than distributed in the desktop
- stateless desktops are those that don't store anything unique and/or important in the VMDK
- stateful is where the VMDK contains unique or important data
- stateful are expensive as it hampers de-duplication (possible with NetApp today - i.e. up to 90% dedupe.  Of course, other de-dupe vendors are available).  VMware just need to work with this technology, no particular actions
- De-dupe is also possible via image optimization with each VDI referring back to a base image and then only having its differences stored as unique files.  Linked-clones delivers this, but some apps can't cope
- Performance can be a bottleneck for VDI (eg boot storms, virus checkers)
- Ideal might be to have the base VMDK images on local storage on the servers and hold the diff files on the shared storage

Data Centre Storage
- what about VMs moving across clouds - ie between datacentres?
- multi site datastore federation - all storage is available in both data centres and VMs can see the data in either location - an extention on top of synchronisation.  Likely to be costly to acquire and manage?
- DR snapshots into the cloud?  Will include the VM status in the VMDK
- User data in the cloud to allow personalization across devices is effectively a back up scenario
-