Tiered Storage – no need to wait
by Hu Yoshida on Jul 3, 2006
The primary value of storage controller based virtualization will come from the ability to move data across tiers of storage without disruption to the application. Virtualization will help you, non disruptively, migrate all those terabytes of storage you bought today to the next generation of storage when their lease expires or you need a technology refresh in 3 to 5 years. It will enable you to buy more of the lower cost tier 2 and 3 storage and still enjoy tier one functionality.
Many people say that they already have tiers of storage without the need for storage virtualization. They use expensive tier one storage for data that needs to be available 7×24 or requires replication for business continuity. They use less expensive tiers of storage for data that can afford maintenance windows and don’t require replication. This type of tiering is really application tiering and not storage tiering. While this is certainly more cost effective than putting everything on tier one storage, all storage based moves and copies of the data for a tier one application continue to reside on tier one storage. Not all the data, especially copies of the data, for tier one applications require the same level of availability and recoverability. Storage controller based virtualization provides the ability to move or copy data to lower tiers of storage without disruption to the host application and without compromising tier one controller functionality like 7×24 availability and business continuity.
Without storage virtualization, host software would be required to move or copy the data to different tiers of storage with some amount of disruption to the application. Host based software would also be limited to the storage that it connected to and would be a single point of failure. Storage controllers like the USP or NSC have no single point of failure as they are composed 128 or 32 processors, respectively, around a shared, write protected cache. Storage controller based virtualization enables the non disruptive movement of less active data or copies of application data to different, even heterogeneous, tiers of storage, without compromising tier one functionality or availability.
Others say that tiering with storage controller based virtualization can only migrate volumes, so it is not practical for migrating less active parts of a file without additional middle ware and extensive work in classifying data which is often a moving target. However, they also tell me that they have 3 to 10 copies of that volume for backup/recovery, data mining, development test, data sharing, archiving, and business continuance. Not all those copies need to be on tier one storage. Without virtualization all snap shots, mirrors, and replication of tier one storage has to go to tier one storage, unless you use host software. A recent post by Steve Duplessie points out in Disk to Disk to Disaster Recover, the roll that lower cost tiers of storage can play in disaster recovery. With storage controller based virtualization we can create non disruptive snap shots of the tier one data volumes to tier two storage, than use a backup product to backup the tier one application data to a VTL solution like Diligent without disruption to the application. You don’t need to do a three year data classification study before you start to enjoy the benefits of virtualization and tiered storage.
As far as data classification goes, there are many applications that follow the classic billing cycle, 30 days, 60 days, and one year before it goes to tape. These can be easily classified for automated or policy based migration.
The CIO for one of our large financial customers plans to reverse the current ratio of 78% expensive tier one storage to 78% tier two and three storage through the use of storage virtulization and storage tiers.
Comments (5 )
So, good comment – even if you’re not considering virtualisation/ILM in general, then consider using multi-tiered storage virtualised for BCV/SI type copies. Yes, this can be done today without virtualisation using different disk types in the same array, but virtualisation enables this to be pushed to a lower tier again. Hu, worth pointing out here that with HDS, the SI copy is an async process so the creation of these copies is not performance impacting on the primary host, enabling cheaper (and consequently less performing storage) to be used for SI data.
There’s a great podcast on the Value of Storage Virtualization over at http://www.evolvingsol.com/sol_storagevirtualization.html
Karen, I see that you have to login before listening to the podcast, is that required?
I usually avoid logins for podcasts, but that’s just me.
There’s a list of other podcasts for folks on the Data Storage wiki, Karen, reach out to me if you want to get some stuff added there.
Actually, you don’t have to login to listen to the podcast. If you look on the right hand side of the page, you can just click on “podcast”.
I see it located in the body text, and it links to the podcast, Sounds like Evolving Solutions has speciality in Storage Virtualization, but he didn’t mention HDS as a focus area!
Send me an email if rounding out virtualization to include HDS products would be of interest –I can connect you with the right folks.
Lastly, if you have an ongoing podcast series, let me know, I’ll add it to the data storage wiki.