Getting On The Bus With Cloud Storage
I've been down on vendors pitching half baked cloud storage visions for some time now, which often seem like attempts to avoid being left out of a feeding frenzy, rather than a vision of how the cloud might improve real storage environments. So I've been fairly negative on the hype of the cloud, and the data seems to support this negativity. According to Forrester's 2009 Enterprise And SMB Hardware Survey, 91% of respondents say they have no tangible plans to move forward with cloud storage. But when I stop to consider it, I don't want to look back in 5 years and say that I was the analyst that said cloud was never going to happen. There's a huge amount of pain in enterprise IT, and storage is often the most problematic segment in the data center. It has ever-increasingly high costs, challenging hiring of the right skill sets, and poor efficiency. Chaos is becoming the norm. The prospect of getting effective and secure services from an outside band of experts or of building out a consistent, measurable, and scalable internal architecture resonates strongly with IT buyers. So, there's big pain, budget funds, and a vague vision of an alternative solution. That's a recipe for eventual change if I've ever seen one.
So why is it that the visions are so vague, and market progress is so slow, in spite of the fact that every vendor is pitching some kind of cloud message and a surprising number of Forrester customers are inquiring about cloud possibilities for their shop? The gap, as I see it, is in the use case. There is no single "Storage Cloud" writ large the way that there is one Internet for all workloads and connections. Just as firms use different physical or logical network fabrics for database, backup, file, management and replication traffic streams, and as they use different array architectures for high performance random read write, sequential file streaming, mid performance application, departmental file sharing, backup and archive data types, cloud network and back end architectures need to be tailored to the workloads they will be used for. What works for one will not work for all.
So while it's tempting to think (and market) that the cloud will eliminate data center storage as we know it in one fell swoop, it's not bloody likely. Instead, buyers and vendors should think carefully about the specific workloads they are currently provisioning for that might be effectively solved in a cloud architecture. Generally speaking, this isn't going to be performance sensitive workloads. While Internet bandwidth has increased, so has data volume and application functionality. Forrester's recently published report "Cloud Storage Comes Down To Earth" explores workloads that lend themselves to the cloud, and points to three that could just work: 1) Whole applications (storage centric cloud computing really), 2) Backup and/or archiving, and 3) File storage. There's no magic in the cloud, and the concept of a consistent architecture with geographic separation from the primary data center is not well suited to all workloads. Know thyself — pick the workloads that make sense and move forward carefully, evaluating the risks and rewards of doing so. And avoid the hype — don't do what’s going to help your vendor validate a vague new business model, do what's going to better solve the significant challenge of managing your firm's most critical asset, its data.