I thought this blog entry by Bob Lazano on Amazon S3 had some interesting insights.
Internet-based storage has been an intriguing topic for a long time. Lately, its re-spun as part of the hype surrounding cloud computing. So, I'll just call it cloud storage here. Fans of cloud storage believe it is an inevitability - just as I think pervasive iSCSI is. They think cloud storage is easy. Detractors don't think cloud storage can provide the reliability and availability required of an infrastructure element. They think cloud storage is gnarly and difficult. I guess it depends on what you are willing to put up with.
Chuck Hollis writes that he thinks the issue is trust. I agree, that the thing that will keep more businesses from using cloud storage is that trust will need to be developed between the cloud vendor and the market. The insinuation is that cloud-trust will carry over from product brand. EMC is certainly one of the more reliable brands in the storage business, but I'm not sure that trusting a brand is going to be the key for cloud storage success.
Chuck correctly points out that scaling is the big issue (yes pun intended). I think the question is whether or not the reliability and availability can be achieved by a cloud storage business. The business has to make money and that either comes from cutting costs or charging enough to make a reasonable profit. But the prevailing expectation for cloud storage is that it is cheap. Cheapness on a large scale is an environment that is ripe for problems. Again, the question is what you are ready to put up with. If you expect something like 5 nines, then I would suggest that S3's problems today cast a different shade of cloud over those expectations.
I don't mean to be unfair to Amazon because I've never seen their facility and I'm sure its a fairly serious undertaking. But that is exactly my point - its the unexpected, strange things that start the chain of events that lead down the slippery slope to failure. There is a big difference between providing application services and primary storage services. Primary storage service expectations are far less forgiving. I think if you were to poll Salesforce.com and Amazon S3 customers about their confidence levels after their respective problems this week, my guess is that the Salesforce.com customers have been less alarmed by the whole thing.
The bottom line question is whether you think you can do better on your own? Today, most storage customers think so and I believe most will continue to think so for a long time, despite what market analyst thinkers like Nick Carr believe. There are no perfect storage vendors, just as there are no perfect storage administrators and no perfect storage software developers, but if you are responsible for steering your IT organization through unforeseen problems, do you think your cloud storage vendor will be as diligent as you? Do you think their best practices are sufficient and consistent enough to protect your data the way you demand? I think most IT professionals want their own fingers on the pulse.
That said, there are storage services that I think can work very well. Most anything that is not primary storage can be done effectively, especially such things as archiving, replication and backup. Secondary storage that protects a company from failure at the primary site or removes culpability for legal & audit risks can be very effective and reduce risk all around.
