VMworld 2010 Reporting: VMware's Vision of Storage

This presentation is a vision of how VMware believes storage will evolve over the next 3 - 5 years.

Future claims

- physical storage will achieve linear scalability and will be managed trivially
- storage will be a single pool allowing vms to move between data centers.

Current problems and challenges with storage

- Planning and sizing over time
- Optimizing HW value
- Cost of managing scale
- Application SLA enforcement
- Multi-tenancy
- Cross datacenter movement

IT's challenge; How can I take a large environment and run it with different SLAs and scale It? How can it be run with zero touch management? Two solutions have traditionally been available; scale out (separate storage tiers) or server and storage combinations (i.e. virtual storage appliances)

In the future storage needs to become vmdk aware and also to understand vm based policies. In addition it needs to support key encryption to deal with multi-tenancy and security.

Enablers for these transitions are multi-site global identities for the vms. Long distance vMotion for site-to-sit load balancing.

Some considerations for desktop and storage architectures include new problems like anti-virus storms brought about by VDI. Challenges include cost, scale (100000 vm deployment should be as easy as the deployment of the 10th vm). Also a state full experience must be delivered to the users.

Storage is moving to cloud based application centered management. Traditional storage solutions prevented applications from scaling. This was helped by scale-out or clustered applications but this involved development coding with architecture in mind and restricted application designs.

VMware believes the storage of the future will be a blob store ( hmmmm not to sure about this concept). Essentially it will lack structure completely opposite to the RDBMS model and instead will be a series of datastore services that will likely be made up of a collection of infrastructures both internal and external to an organization. The benefit to this architecture is low OpEx as there is no structure to manage.



- Posted using BlogPress from my iPad