The bigger issue with UCS is vendor lock-in, which offers pros and cons for the organization. Aligning with a single vendor can provide benefits such as a single source for assistance and support. But there can be a tremendous downside if the particular solution impedes key business initiatives or growth, fails to solve the underlying technical requirements that you were looking for, or prevents interoperating with other important non-UCS...
equipment or software still operating within your environment.
The last point of interoperability with non-UCS technology cannot be overstated. For example, a UCS platform may be designed to run VMware hypervisors, but if you aren't able to switch to Xen or Hyper-V in the future, that might be a problem looming on the horizon -- limiting your data center's future flexibility. It's a similar concern from a hardware perspective.
"If I bring in a Cisco [UCS] and I still have Juniper and I still have other equipment in my environment…and I have to tie all these together, how well are they going to play?" Schulz said.
Then, there's the question of the vendor itself. Choosing a UCS means committing to a vendor and its roadmap, but vendors change direction with their product lines, drop product lines, merge product lines, change their emphasis on product development and support, and get acquired by other vendors that may not adhere to the planned roadmap.
Consequently, a move to UCS should involve a serious examination of the vendor, their stability, the stability of their partnerships (such as Cisco's UCS work with VMware), their development roadmap, and consider a contingency in the event that a UCS platform becomes untenable. Perform due diligence in cost analysis, testing and evaluation, and a thorough analysis of the vendor agreement to see any restrictions that might present potential business problems down the road.
UCS includes servers, storage, and networking elements, so deployment will typically involve an exchange of most (if not all) of those sub-systems to those provided by the UCS vendor. Beyond the hardware, UCS deployments will also include a hypervisor and virtual machines (VMs) along with data management and data protection tools. But, the actual implementation typically starts with upgrades to the network -- both in the LAN and the SAN.
"We put into production a pair of Cisco 4507 switches to give us a 10 gigabit infrastructure to support the UCS," said Rod Gabriel, IT infrastructure engineer with United Financial Services in Grafton, WI. "We also went live with a pair of MDS 9124 switches to give us the SAN switching infrastructure that we needed."
Additional storage capacity may also be needed in preparation for UCS, but that need is most notable when UCS adoption is coupled with a significant increase in computing capacity -- such as bringing more host servers and VMs online.
Once the groundwork has been prepared, the UCS system (such as the Cisco UCS) can be deployed along with new servers. But the actual transition to UCS is not immediate, and experts expect to take about one month to test, configure and gain experience on the new platform.
"About a month out from there [installation], I expect to deploy new vSphere hosts behind UCS," Gabriel said. "Once there, I'll migrate my entire virtualization infrastructure over to UCS, and by sometime in April be off of all of our existing ESX servers and fully onto the UCS platform."
The introduction of a complete UCS platform should be reasonably problem-free, but there are several post-deployment items that prospective adopters should consider. First, keeping up with server and storage growth can be a challenge for some organizations. Experts report that UCS accommodates the organization's growth demands well, but it has renewed a focus on capacity planning practices.
However, Heil encountered an issue when deploying an operating system other than VMware ESX 4 and Windows 2008, and he strongly advises prospective UCS adopters to plan operating system requirements thoughtfully and test the versions that you expect to use.
"We didn't bother understanding how to roll out something like Windows 2003," he said. "We ended up having that need to load [Windows] 2003 on a blade, and from a driver perspective, it was a challenge and required a ticket with Cisco to get the process down correctly."
Another potential bane lurks in the need for periodic code updates for the UCS platform. Since UCS is a complete system, a code upgrade glitch can impact the entire environment. Administrators should take extra precautions to protect data and have contingencies in place before executing on UCS code upgrades.