I was at the Enterprise Cloud Summit in New York where Lew Tucker was moderating a panel on private clouds in the enterprise, and I noticed a rather dramatic change in the discussions that took place regarding private clouds. To begin with, Alistair Croll from bitcurrent gave a wonderful set of opening remarks that framed private clouds in a way that is really accessible. Randy Bias talked about the experiences with his new company, Cloudscaling.com in architecting a cloud for Korea Telecom. Again, much of what he had to say resonated with our experiences at Eucalyptus (perhaps too well at times) -- a great talk. The morning session ended with our panel covering topics offered by Lew at first, but mostly from the engaged and surprisingly awake audience members.
What struck me during the break, however, was how much emphasis both the talks and the audience put on the subject of manageability, and conversely how little focus there was on the topic of scale. Lew even encouraged discussion of scaling issues, with little effect. It seemed to me that the primary concern voiced by both speakers and audience pertained to capabilities for users and administrators to manage a private cloud.
This shift in concern, if it is indeed taking place, is both necessary and important. Clouds (private and public) are capable of doing many tasks simultaneously and at machine speeds that are otherwise sequential activities usually performed by people. If a person can create a problem using a conventional infrastructure, a person using a cloud can create it faster and with wider reaching impact if the appropriate safeguards are not in place. Developing the interlocking tool chains required to prevent the outbreak of cloud-based chaos is what cloud management companies such as RightScale, rPath, and New Scale do well, but the question of what policies these tools should implement remains, for the moment, largely unaddressed.
That so many of the audience members were concerned with cloud management policy is (to me anyway) a sign that enterprise cloud deployments are maturing. During the last year, most of our customers have been focused on understanding and becoming facile with cloud technologies. Now, it seems, they are moving beyond the question of what components are needed for an enterprise cloud to questions of how IT process should be customized for specific use cases. Process development for enterprise clouds is certainly challenging, as was process development for data center virtualization before cloud, and process development for Linux deployment before data center virtualization. For enterprise clouds to become as ubiquitous as many have predicted, the shift in focus from technology to process and policy development must occur. My observation from the Enterprise Cloud Summit is that this important developmental phase is now underway.