Recently I got a Celerra NX4 storage array to meet my organizations storage needs, or out of the box, solve a specific problem that we are having with regard to storage. Slow data performance across the network and Windows Update. I found out quickly by doing some simple math covering what exists today and the maximum amount of available storage on the NX4 (~900GB) that this move to SAN Storage would indeed be something that has multiple phases (read disk shelves).
It is not to say that there was no planning ahead, we did also purchase some HP Proliant DL 380 G7s for the upcoming virtualization of server workloads, but given the amount of disk we are consuming just for files (non-deduped files at this point) the opportunity to tackle everything at once is thankfully in no way possible.
Data growth will happen
I like to think I am a realist when it comes to my job. Sure it is cool to have the latest greatest infrastructure, but if the gear in place today is getting the job done why change? We are storing our information on a Windows Storage Server appliance with direct attached disk configured for hardware RAID 5. The useable storage available in total is about 675 GB, which isnt too bad considering that we have used about half of that disk space over the five year life of the device.
Some users, consume way more than they should and all have a surely plausible business reason for their usage. Because there havent been huge spikes in the growth in other areas, this is ok. Until the data begins to saunter across the Ethernet to the NX4. When it arrives, there will be limits in place for the user and deduping going on to help the limits be realistic.
Sure we went to the current storage server because I needed some type of disk space in a hurry on the cheap, but it was admittedly poorly thought out because it requires forklifting to grow the storage. Because more data is being consumed and many end users assume that IT is in charge of what they consume, many things, even those from 2001 need never die. While this might be true, it isnt practical, especially in a small business. Keeping old data forever because it has never been thought to clean it out is just bad management. Both by me, the storage admin, and the user.
Because of this, capping or quotas or <insert limit buzz word here> will happen when we transition. Sure there will be exceptions, but certainly not for backups of itunes libraries and photo albums.
Figuring out what is reasonable
Determining accurate growth certainly is a science and plain dumb luck. Because we have had our current file server for five or so years and have just passed the half way mark 20% growth is more than reasonable. I didnt hire a crack team of scientists to determine this, I purely looked at current use versus total available to see what would be left over. Then tried to find a target quota amount for the users home directories.
Some users will be giving up some disk space, but I think this is good for everyone. Sure we can get more shelves and we likely will in the future, but slowing the growth to a trickle is something that will hopefully get people to be more mindful of what they keep. If the newest Justin Bieber CD in the itunes library wont fit because you havent the space to keep it, should you cull your library or store you collection elsewhere, or remove data needed to do your job? I think the choice is obvious and can recommend great small form factor disks for iTunes storage, but it is ultimately up to the user (or their supervisor) to ensure they store the correct information.
I think the move to scalable storage was definitely correct for this situation because data needs will only get more complex as the world moves forward and with virtualization becoming more and more popular, the savings brought about by “less with more” or at least less compute space for more compute power will be very well worth the slight bit of upfront and transition costs incurred by SMB.
Changing over to SAN storage was a very big shift for the business side of the house, especially since they have not started using this hot new thing as of yet. When planning our storage needs, I was concerned with growth, but included server consolidation to reduce the amount of servers (read “four letter words”) that were in the data center (back room closet). Doing this seemed to appeal to everyone, even though we are quite a ways out from actually virtualizing any workloads.
What now? The discussion begins
There was a lot of prep work that went into getting the new equipment ordered and there will be a lot more work in ensuring it works to meet the needs of the users now and in the future. I expect it will grow considerably much sooner than the business side will anticipate, but when it does, the long term savings on power and other areas will be well worth the initial investment.
Implementing guidelines to help users make better decisions about what they keep and where they keep it might be a good move, at least for my organization. It will help keep things that need to be kept and remove things (or store them for less time) that are not as important. Some times things will get done to ensure (or at least try to ensure) timed growth, like quotas and deduping, and other things will be done with the direct involvement of the user and place the responsibility on them.
The best thing I have learned in preparing this new storage is that there is a good deal of work and consideration to do before letting the users loose on the thing once I get that far, or just before… I will have more to say here.
