Tuesday, June 24, 2008

Postcards from Gartner's IT Infrastructure, Operations & Management Summit 2008 (Day 2)

Today's presentations were almost entirely about virtualization with generous servings of analysis of cloud computing.

Thomas Bittman
Thomas Bittman opened with a keynote on Gartner's predictions on cloud computing, and the likely march the industry will take over the next 5+ years toward this eventuality.

Gartner's predictions are that there will be many varieties of cloud computing, from the AWS-style of raw hardware, up through various types of service providers of platforms, services, and even component services that will be wired-together by other types of providers.

Bittman even went so far as to suggest that service "brokers" could emerge. For example, you've established an SLA with a cloud computing service provider, and for whatever reason, that SLA isn't met (maybe AWS has another glitch). Instantaneously, your broker finds another compatible cloud and "fails-over" instantly to that provider.

Gartner's sense was that there will likely be a few "mega" providers (AWS, Salesforce, Google, MSFT, others) and then hundreds/thousands of smaller mid-market and specialty providers... not unlike the evolution of the hardware market today. And on that note, they also predicted that the hardware providers (like Dell) will probably get into the hardware-as-a-service market shortly as well. That should be interesting to watch.


Cameron Haight
Next, Cameron Haight spoke about emerging "virtualization standards."

He made the very reasonable assumption that users will want to manage multiple VM technologies using a single tool. (And, with a straw-poll, the audience conclusively agreed).

A few of the initiatives already underway include:

* DMTF (Distributed Management Task Force) is already working on draft specifications for interesting standards... not for VMs, but for properaties that would aid in management -- such as a Virtual system profile (i.e. for re-creating a set of VMs), and a resource allocation capability profile (i.e. for monitoring managing VM hardware resources like CPU, memory, network ports, storage, etc.

* also an Open Virtualization Format (OVF) is underway. This isn't a standard for VM files. Rather, this would tag VMs with metadata to ID them, say for packaging/distribution. For example, it would help characterize what's "inside" a VM before powering it on. My suspicion is that this could be the foundation for a "common" type of SW container, and a common approach to monitoring/managing such VMs. But I also suspect that the vendors will either (a) fight this tooth-and-nail, or (b) adopt it, but "extend" it to suit their needs...

Cameron also ran a few interesting audience polls during his session. Follows are some notes I took, but I believe he'll probably publish them in a forthcoming research note:

Q: What VMware products are you currently using?
71% VirtualCenter
36% Update manager
31% Site recovery manager
23% Lab manager

Q: What do you think is most important for VMware to focus on?
30% optimization of VM performance
18% VM sprawl
16% Maintanance/patching
12% Accounting/chargeback
11% Root-cause analysis

Yep, these pass my sanity check. It should be quite interesting to see what VMware's next moves will be.

No comments: