David Grant is the data center manager at Mitel Networks Corp., a telecom based in Kanata, Ontario. He started with data deduplication in 2008, and would like to see more innovation from vendors for capacity management.
SearchStorage.com: First, a look back at 2008. What was your biggest storage project that was completed this year? What problem did it solve for you? What do you wish had gone differently?
Grant: There was a lot of planning and what if's, but most significant was dipping our toes in data deduplication with a Quantum Corp. DXi appliance. We hoped to reduce our backup overhead and improve service for faster restores with more online capability. The results were mixed, but due more to [a] lack of internal resources to focus on the initiative.
SearchStorage.com: Can you elaborate? Was there not enough configuration or tuning time on the box?
Grant: We tried to run to a deployment too fast and needed to backtrack, but found that difficult to do once we had live backups on the device. It was a combination of not fully understanding what we were doing in the initial configuration, and then not realizing the impact of trying to undo and correct it. The Quantum team has been very supportive and they've tried their best to help us, but we're not exactly following their best practice guidelines so we have to accept a large portion of the blame for our difficulties here.
SearchStorage.com: What effect will the global economic climate have on your storage planning for 2009?
Grant: Little impact, as we were already well advanced in planning and were given approval to proceed recently.
SearchStorage.com: What's your most important storage project for 2009?
Grant: Data center consolidation of our Phoenix site to our Kanata facility, involving a 60 TB storage footprint in Phoenix, two new storage frames plus an upgrade to an existing one, data migration and rework on SAN. And we're planning to move servers--as data increases, so too does the impact on storage.
SearchStorage.com: What storage technologies are you evaluating for 2009?
Grant: I'm very interested in data deduplication for primary storage. I need to rein in my [VMware] ESX storage and it's an ideal candidate for this technology. It also offers hope of implementing an affordable ILM [information lifecycle management] architecture at long last.
SearchStorage.com: How does dedupe for primary storage impact ILM architecture?
Grant: Imagine a storage frame with mixed spindle types ranging from 72 GB 15,000 rpm to 450 GB 10,000 rpm and 1 TB FATA, add in a VTL [virtual tape library] device and then layer over a data dedupe appliance. Add in data deduplication with some policy-based storage management features, and you have an ideal ILM environment.
Critical data--usually created, updated and read frequently for two weeks to four weeks--gets stored on high-speed native. Older data is automatically transitioned to midtier spindles and possibly deduped until no longer active, as in only needed for roll ups and reporting, occasionally referenced. Then the inactive bulk of data that needs to be available at short notice for reference and lookup gets moved to lower tier spindles and [is] fully deduped. Lastly, data with a long-term retention requirement for compliance gets transitioned to a VTL-type device or alternative.
Now imagine all this being done under a single data deduplication algorithm and policy-based engine, and the volumes of data you're moving around don't need to be un-deduped until required by an end user. The result should be a smaller storage footprint; data migrations should be very fast, as we're handling less bits and bytes and, consequently, it should take less processing power on our ILM platform. I believe that all the necessary pieces are out there and being worked on individually, but is someone working on gluing them all together?
SearchStorage.com: Is there any other technology that would be useful for accomplishing your next goals?
Grant: A close second would be a common standard for VM [virtual machine] formats. With an increasing number of players now touting virtualization offerings, people like me are being offered multiple choices in bringing virtualization into our data centers. But with the current state of play, if we make a wrong choice, it's a rip out and replace, which is very expensive in most cases. If vendors agreed on a common VM format or opened up to one another, I might be able to change vendors without having to re-virtualize all of my existing machines and probably storage.
SearchStorage.com: What would you like to see happen in the storage market in 2009?
Grant: An understanding by storage vendors that we buyers can no longer justify the continual buying of additional storage capacity to be consumed by stale data. We need their help in managing that growth and minimizing the costs of retaining data in any recoverable format.