I’ve found that having different tiers of storage is rather common - but having intelligent tiers is somewhat rare. Storage virtualization actually enables you to implement “intelligent” tiered storage in your data center. Based on my customer interactions - 60-70% (some customers even have 80%) of all data isn’t accessed 90 days after its initial creation. If you have 100 TB of data - that means that 60-70 TB isn’t being used. That’s pretty staggering.
Storage virtualization provides a logical layer of management and control enabling data to be moved seamlessly between different storage systems. Implementing intelligent storage tiers is an extremely effective way to leverage what you already have on the floor and free up valuable capacity on your Tier One storage. You can move that 60-70% of dormant data off of your most expensive storage and put it on lower storage tiers. The beauty of using this approach is that it is fundamentally transparent to users and applications. The data remains online and no remapping is required. If you ever need to access that data again - there is no change in operations - it is there.
Take a look at David Merrill’s analysis of tiered storage - http://www.hds.com/assets/pdf/tiered-storage-economics-defining-and-calculating-the-economic-benefit-of-tiered-storage-solutions-wp.pdf. The report gives detailed analysis for implementing intelligent tiered storage. I believe that David is being conservative claiming you can get a 15% total cost of ownership (TCO) reduction but it can be much more than that. In the report he provides well thoughout examples in which hundreds of thousands of dollars (even millions depending on the size of your environment) can be saved using storage virtualization and an intelligent tiered storage strategy.
The point of the last three storage economy blogs is to present some ideas on how to leverage storage virtualizaiton that can have a real economic impact. I’m sure you’ve read a bunch of things about utilization, thin provisioning and intelligent tiering - these things actually do work and can significantly reduce cost. This is a call to action - if you will - for you to consider making them priority projects in an arguably unprecedented economic climate.
Comments (2 )
Michael Hay on 27 Nov 2008 at 8:02 pm
Tony, good read and salient points all. Having been there when we built part of the management stack and modified the underlying migration engine within the USP(V) systems it is good to see these things being recognized. One of the goals that we had in mind for tiered storage deployments was making it drop dead simple to create tiers to your heart’s contentment. So we looked at highly flexible models for expressing tiers and realized that what Apple did with iTunes in the form of smart playlists we should do with tiers. So a user can easily build a tier which is all S-ATA and at an ArrayGroupBusy percentage of less than 50% and then quickly migrate their data to that location all without having to reconfigure the servers and applications. Further there is a little “candy” we put into the mix which is caring about data safety. Notably, when a migration is performed the user can either click a check box to perform a simple write of 0’s over the source once the migration is complete, or if they are really paranoid the user can issue a shredding command via the engine with in the USP(V) by CLI. All of these kinds of things are critical when considering an array retirement use case, and I think it speaks to both the intelligence that you talk about and care for our users’ data that Hitachi always has.
Thanks,
Michael
tony on 01 Dec 2008 at 9:48 am
Great feedback. I strongly recommend that HDS put together a ton of use cases giving details of real world implementations.