I was at a company recently that had 60 TB of NetApp SAN Storage, in 2 different data centers (primary and business continuity). When they put it in 3 years ago, there wasn't really a compelling cloud storage alternative. But now there is, so the question arises - where should the next 10 TB come from?
These 60 TB cost $1.5M in hardware, software and implementation fees. I don't think it's reasonable to capitalize any IT gear for more than 3 years, so let's call it $488,260 a year in capital depreciation. The service agreement runs $174K for 3 years. So one year costs come to $8.81 per GB. Another way to look at it is $106 per month per user for data storage provisioning.
ouch.
Is NetApp local storage "better"? Absolutely. It's faster and located next to servers, with servers connecting to the backplane at 100Gbps. You need high-speed connections for databases and transactional systems, for data warehouse ETL, and query. For VMWare images to properly replicate and backup, local connection and company WAN speeds are critically important.
But all of those VMWare snap and mirror images - they're huge. And that's another drawback to the SAN - you need to duplicate storage for disaster recovery. So it wouldn't be a stretch to say that they're really spending $8.81 x 2 = $17.62 per GB per year. There are other costs to consider, some of which are small, but they all add up. Take power to the SAN. Let's say the whole thing draws 6KW. That's another $5,256 per year just to power the SANs.
The much bigger cost is, of course, staff to manage the SAN. I haven't developed benchmarks about how many systems engineers it takes to manage a given size and complexity of SAN, but for our example let's say it adds $100K of resource burden, annually. You can't get away with zero IT staff managing cloud storage, but it is much more of a vendor management role than a systems engineering one.
So NetApp is better, but is it 4 or 8 or 10 times better? Well, for user storage at least, the answer is an emphatic no. In fact, something like Jungle Disk, Box, Dropbox or, to go a more secure route, Amazon S3, offers much more in the way of functionality and anywhere, anytime access for the end user. It's easier to charge back. If they need another terabyte today, no problem. So end users love cloud storage, in contrast to the traditional IT department that puts 100MB limits on them and takes weeks or months to add additional storage.
When Windows first hit the market, it was competing against Novell for file server market share. Novell was in the 75% market share range. So IT managers like me purchased Windows servers and put them alongside the Novell servers in our data centers. More and more services started being ported to the Windows servers, and we all woke up one day and said well, Novell may be a superior fileserver, but I don't need it anymore. I can get adequate file services from Windows plus a whole lot more. So now, Novell is no more and Windows dominates the corporate data center.
I can see a similar path for Cloud Storage versus SAN storage. Companies will start to put their next 10 TB of storage in the cloud. When that works out, they'll move more and more there. At some point, IT managers will start to question the investment in local SAN storage, and they will pressure their vendors to make moving to cloud storage a workable proposition for even heavily transactional and high-traffic data.
The real death knell will only toll when we no longer have our own data centers. But I don't think that day is too far off. Many companies I have consulted at are in a transition similar to Novell->Windows. They have outsourced many of their critical applications. So they may see the day when everything is outsourced. In the meantime, it would pay to look at storage costs and alternatives.
Interesting piece. It is not necessarily a cost saving but investment may shift from storage to network to meet cloud demands and resilience. I would love to see a measured response to your artilce from an on-site storage advocate.
ReplyDelete