United States
Site Map Contacts Hitachi Global Community
Hu's Blog - Data Storage and Virtualization Thought Leader Hitachi - Inspire the Next

Hu Yoshida's Blog - Vice President | Chief Technology Officer

Home > Corporate > HDS Blogs > HDS Bloggers > Hu's Blog
Products, Solutions and more

Hu's Blog

Storage Virtualization, Tiered Storage and Dynamic Provisioning

by Hu Yoshida on Aug 19, 2009

Tiered Storage is one of the best ways to reduce storage costs if it can be done dynamically. By dynamically I mean the ability to move data between different cost tiers of storage without disruption to the application. This type of tiering requires storage virtualization. Without storage virtualization, the movement of data between tiers requires the application to be stopped or suspended while the data is moved or copied between tiers with software running in the application host or in an appliance that is connected to the tiers of storage.

With storage virtualization, the application can continue to run on a virtual image of the data while the virtualization engine moves or copies the physical data between the tiers of storage. With the USP V virtualization, where the virtualization is done in the controller cache, the movement of data between tiers is done by changing pointers in a separate control store and the cache image from one tier is redirected to the internal or external array group of the other tier. All this is done in the USP V and there is no impact to the SAN or to the application interface port processor. The interfaces to the two tiers of storage are driven by separate processors in the USP V.

The difference between copy and move should be noted. A move is the transfer of data from one tier to the other. Today that transfer for blocks is done on a volume basis. For filers like our HNAS product, the movement can be done by files. Movement of data can take time if the volumes or files are large. While the popular view of tiered storage is that data will be moved up and down tiers based upon changes in life cycle, in reality, movement of data is done only for major changes in service level requirements.  Frequent movement of data between tiers of storage also requires the reservation of capacity on these tiers. In other words moving data between tiers of storage on a frequent basis is not practical. With the ease of data movement provided by virtualization, one can simply assign data to tier 2 and only move it to tier 1 if it does not meet its service level requirement or move it to tier 3 if it becomes inactive.

Copying of data between tiers is more practical. In order to keep applications running 7 by 24, we make copies of the primary data which we use for backup, development test, data mining, data sharing, extract, translate, and load or other offline processing requirements. Instead of interrupting the application to make the copies, the storage control unit makes these copies, and with virtualization these copies can be made to a lower cost tier or, in the case of data mining for business intelligence, to a higher performance tier of storage. Copies may require some suspension to the application in order to get a point in time copy or a transaction consistent copy through application interfaces like VSS.

In most cases, tier 2 or tier 3 storage behind a USP V with its large global cache and high availability architecture will see improved performance and availability and will have access to all the enterprise services of the USP V like, replication and dynamic provisioning. The virtualized tier 2 or 3 storage will have the tier 1 services of the USP V without the tier 1 cost. If copies and moves are combined with Dynamic Provisioning (thin provisioning) pools, than they will only move the pages of storage that are actually used and eliminate the over allocated waste that is common to most open systems volumes.

By using storage virtualization to provide dynamic tiered storage you can reduce the amount you spend on tier 1 storage. And with Dynamic Provisioning you can reclaim the waste of allocated unused capacity, and reduce the time to copy and move over allocated volumes. In many cases we see a 25% reduction in TCO/TB/Year. For more information go to the storage economics website or check out David Merrill’s blog.

If you happen to be in Beijing, you can catch David’s Keynote at the China Users Conference in Beijing next week. August 25.

Related Posts Plugin for WordPress, Blogger...

Comments (4 )

[...] not be too quick to throw out the baby with the bathwater. As Hu Yoshida has discussed in a recent post, tiering has come a long way in light of new technologies, particularly virtualization. In our [...]

Barry Whyte on 21 Aug 2009 at 2:11 pm

But you still have the very high tier1 cost of getting the USP-V in there in the first place do you not?!

What if you could make tier2 storage perform like tier1 at a fraction of the cost? And not need to buy a big expensive monolith to gain that… You know what I’m saying :)

Hu Yoshida on 24 Aug 2009 at 8:04 pm

Barry, you make a good point. If you can make a tier 2 storage (which I believe you mean XIV) perform like a tier 1 why would you need a DS8000?
I believe IBM is still selling the DS8000.

Barry Whyte on 27 Aug 2009 at 1:33 pm

I was meaning mid-range boxes like DS5000, EVA, etc when combined with an appliance like SVC or USP-VM. DS8K and USP can of course attach to IBM System Z natively, where most modular systems don’t.

Hu Yoshida - Storage Virtualization Thought LeaderMust-read IT Blog

Hu Yoshida
Vice President and Chief Technology Officer

Connect with Us


Switch to our mobile site