Unified Fabric: Sooner Than You Think
After last week in VMworld, I am convinced that we will see broader adoption of a unified fabric in the server access layer sooner than later. For starters, the technologies that were introduced earlier this year are emerging from the certification cycle which is significantly increases the comfort level for customers. For example, Emulex, Intel, and QLogic adaptors have been added to the VMware ESX Hardware Compatibility List and EMC is expected certify FCoE solutions within the next few weeks. More importantly, I think VMotion and its variants will create a compelling driver in the data center. Experience with customers indicates that they are looking to take broader advantage of VMotion and automated variants such as DRS. Coming out of VMworld, it is clear that VMotion and virtual machine portability will play a central role in the future evolution of server virtualization and the virtualized data cener. In this environment, a unified fabric becomes a critical piece of the puzzle, because a unified fabric ensures that consistent and ubiquitous network and storage services are available to every network connected server. With a unified fabric, you can truly move a workload around your data center without worrying if the destination server has the proper I/O profile. If you are using automated tools such as DRS, then consistent and ubiquitous I/O capabilities aren’t just nice to have, they are mandatory.A few weeks ago, Mario Apicella had a post on some of the benefits he saw of running VMs in a unified fabric environment
“VMotion made moving a VM from one server to another as easy as dragging and dropping, provided that all other conditions were met. FCoE devices such as the Emulex CNA and the Nexus 5000 provide that level of network virtualization that removes most of the obstacles to a smooth VMotion. It’s a match made in admin heaven.”
In a recent post, Rich Miller noted the prominence of I/O at VMworld:
“I heard and saw a lot of backup and DR. Storage is big, and the converged I/O required to make it sing within the virtualized datacenter made a pretty significant showing at VMworld 2008.”
“…networking of the virtualized datacenter really got its props…there seems to be an appreciation for the importance and complexity of “getting the networks right.” — access networks and storage networks…”
My own experiences at the show mirrored Rich’s. Whether it was during sessions or in conversation, I started to see the light bulbs light up as folks started to the understand how a unified fabric could do more than just reduce TCO, it could create a uniform I/O environment in their data center in which there VMs could freely frolic and roam to their heart’s content.