Cloud Connect is a leading conference that brings together much of the cognoscenti of cloud computing. Google's director of engineering, Peter Magnusson; Cloudscaling CTO and prolific blogger Randy Bias; Zynga CTO Allan Leinwand; Rackspace Cloud founder and chair of the OpenStack Project, Jonathan Bryce; and Geva Perry, author of "Thinking Out Cloud".
But even so, I suspect these and additional distinguished participants will not have all the answers, try as they might. Some of this is due to the extremely youthful nature of cloud computing, where choices are proliferating perhaps faster than we can rationalize them. And in some instances, answers are hard to come by because vendors want it that way.
Here's some things I wish I could see at Cloud Connect, but don't expect to:
1. -- A universal translator of virtual machine files, so that they move smoothly from a cloud with one virtual machine format to one with a different format. No human tinkering or intervention required. Failing an advanced, StarTrek-like translator, I'd settle for an agreement among providers on a shared virtual machine format. But don't hold your breath. Microsoft is starting to manage ESX Server files from VMware under Virtual Machine Manager as well as Hyper-V and Citrix XenServer but will it embrace open source KVM from Red Hat? Hmmm. Not this year. Amazon maintains a Klingon-style belligerence toward all file formats other than its own Amazon Machine Images; in other words, it doesn't want to recognize outsiders. An ability to translate a given file into a new format and back again is not rocket science. Compared to the problems already solved, getting to a shared file format is a no-brainer, but it's not in the vendor's interests to collapse boundaries and make workloads-without-borders a reality.
2. -- Common units of measure for cloud pricing. What is an entry level server on Rackspace compared to GoGrid, Amazon EC2, Google App Engine, and Windows Azure? Can I compare the compute power of one to the other per dollar spent? It's not easy. How about bandwidth and memory charges? They're hard to compare when the virtual resource combinations are mixed up in various ways. Funny how it's hard to compare pricing.
3. -- Cloud workload performance monitoring. I have some visibility into the cloud, but when it comes to seeing my own workload, I consider the cloud translucent, not transparent. If I'm lucky, I can see a faint reflection of it, which tells me whether it's up and running. I can't tell how well it's running, unless I test it constantly myself. Will performance monitoring be a standard cloud service someday? In some cases, you can add your own performance metrics, for a fee.
4. -- What about those noisy neighbors, who keep my virtual CPU grinding away on idle as it waits for access to an I/O port? Who invited them? I'm losing1%, 2% or 2.3% of CPU output in the periods of peak activity for my cloud host, and I can't do anything about it. I'm not supposed to know who they are and I'm not supposed to move out of the neighborhood just because they moved in. But some dark night, when no one's looking, I'm taking my files and fleeing to a new host in different availability zone. Don't ask why me why I'm moving. I just am.
Cloud computing is about transparency, about efficiencies of scale, about capitalizing on elasticity. It's also about granting resources to users when they need them, charging them only for what they use. Are we there yet? I plan to ask at Cloud Connect. But I urge you, don't hold your breath.
Nominate your company for the 2012 InformationWeek 500--our 24rd annual ranking of the nation's very best business technology innovators. Deadline is April 27. Organizations with $250 million or more in revenue may apply for the 2012 InformationWeek 500 now.