The announcement this week of the launch of Amazon's Elastic Block Store (EBS) has added another vital piece to the overall cloud computing picture. The EBS announcement is particularly significant since it takes the gloves off when it comes to meeting the demanding needs of enterprise class computing requirements. We can now source our software, computing, storage, and applications where it makes the most overall sense.The Elastic Block Store finally makes it practical, cost effective, and relatively easy to put traditional storage and processing of very large amounts of data in the cloud from a credible vendor.
While Amazon has been offering extremely inexpensive, highly flexible, on-demand computing power over the Internet for several years, up until now their Elastic Compute Cloud (EC2) had a maximum associated storage capacity of less than two terabytes or required the use of their powerful but non-standard storage services, either S3 or SimpleDB. More significantly, though Elastic Compute cloud instances had decent amounts of storage for some organizations, they could not load up and run in a practical amount of time, on-demand. The Elastic Block Store now makes it possible to have dozens, and potentially hundreds, of terabytes of readily-accessible persistent network storage in a traditional format of choice, particularly relational databases, all at commodity prices.
The Elastic Block service also considers reliability and backups first class citizens (as do enterprises) and effectively offers all storage in a replicated, RAID-like environment. EBS also has an innovative snapshot capability that allows block storage to be imaged quickly onto S3, so that a time-sequenced set of backups can be created and made available whenever needed for restores.
Amazon's CTO, the esteemed Werner Vogels, took a deep dive into the Elastic Block Service on his blog this week.
Economies of Scale: The New Cloud Commodity
Amazon has been consistent about the purpose of its infrastructure Web services from the beginning; they are committed to redefining the economics of computing by using their massive infrastructure to achieve best-of-breed economy of scale. In this, they have been quite successful and the numbers tell the story...
The new Elastic Block Service has a similar pricing model to Amazon's other infrastructure services and you pay only for what is used. At 10 cents per gigabyte per month for storage and 10 cents per million I/Os (cached reads/writes are essentially free), 10 TB of completely utilized storage (i.e. filled with business data) with round the clock use by 5,000 enterprise users performing an average of 10,000 I/O requests per hour would result in a bill of only $4,600 per month. This is a remarkable figure and shows the economies of scale that Amazon is prepared to offer to organizations willing to leverage it.
The advent of a true commodity cloud computing solution that allows the operation of traditional computing applications of choice, a capable backup strategy, major economic advantages, combined with a vendor that has proven stability and a long-term future, and the result is a recipe for serious change in how enterprise computing can be achieved for many organizations.
Jeff Schneider, CEO of SOA firm MomentumSI, posited to me today that recent developments make the writing on the wall clear: We might just be ready to declare the "time of death" for the enterprise data center, and I'm hard pressed to disagree, even if it will take most organizations 2-5 years to realize it.
Now, don't get me wrong, I don't expect most organizations to move wholesale to services like EC2 + EBS immediately, far from it. For one thing, enterprises are notoriously slow moving and risk averse when it comes to major IT changes. And certainly there will be a great deal of early skepticism from many quarters about such a strategic shifts, some of which will be valid and much of which will not be. But it's fairly clear that the classical multi-hundred thousand square foot proprietary data center is a dinosaur of another age, like mainframes are for most organizations today.
Objecting to Outsourcing the Data Center
Though most businesses are quite comfortable in using external utility services for electricity, water, and Internet access -- and we even use banks to hold and pool our money with others "offsite" -- we are still largely unready to move computing off-premises, no matter what the advantages. The classic challenges stand in the way, including proving out how to secure, manage, trust, govern, and fully exploit the benefits that cloud computing can offer. Worse, we have to do the hardest thing of all, change the way we think and work. Fortunately, the advent of Software as a Service (SaaS) has helped pave the way for Platform-as-a-Service and we're frequently readier than we think.
One of the core principles of Web 2.0 is "Software Above the Level of a Single Device." This has taught us the strategic value of 3rd party open APIs, widgets, and many other forms of 2.0 era distributed computing in the cloud. Today, the best applications tend to use resources from the best source, wherever it lies on the network. Whether that's a Google Map widget using Google's vast server farm and location data or a SugarCRM instance running on EC2 and EBS matters not. Increasingly, we can now source our software, computing, storage, and applications where it makes the most overall sense.
For several years now, business leaders have wanted to open up their organizations strategically to the cloud. Cloud computing can theoretically make make this more manageable and less impactful on the overall enterprise by providing access to almost unlimited on-demand, partitioned resources so that SOAs can be more federated, distributed, and scalable.
And at this point in the very early stages of the cloud computing era, that's probably the best way to look at it. Because I'm not advocating that we mindless push all our computing requirements out to the cloud to reap simple economic benefits, the discussion is far more strategic than that. However, we do owe it to ourselves and our businesses to seriously consider the benefits of locating our computing environments in the right place where it makes the most sense with a full sense of the various considerations that impact us. When we do this, however, the equation will inevitably balance out more and more towards outsourcing much of our data centers.
Core IT applications such as HRM, CRM, and ERP as well as line-of-business application may seem like the last that will go to a cloud model. But while they certainly won't be the first, they may actually not be far behind when IT managers weigh the total overal benefits, particularly during an upgrade cycle. If there will be a lag, it will likely be in the upper tiers of the Fortune 500, where the sheer size and complexity of the IT landscape will impose it's own delays. Thus, the traditional data center won't disappear overnight, but it will almost certainly shrink on a regular basis from now on.
Finally, it should be pointed out that software products are often years ahead of the market. This is no different with cloud computing and many vendors have actually been in the space for years now. But the relentless forces of commoditization and competition are having their say as well and cloud computing offers up very substantial bottom-line returns. Throw in an economic downturn and a round of enterprise cost-cutting and the market and cloud computing seem ready to meet.
Be sure to read Enterprise Cloud Computing Gathers Steam to see how the cloud computing/PaaS space is rapidly filling out from a vendor perspective.
Are you looking at cloud computing for your enterprise? Why or why not? And what are you planning to do?