Request For Information: Data Center Networking

In the coming months, I am going to be spending a lot of time getting up close and personal with networking vendors' data center product sets. Jim Metzler and I are working on a data center networking RFI that Network Computing will publish, along with the vendor responses, in October. Jim and I are also going to present the results at the Interop NY show in October. I am looking forward to developing the RFI and reviewing the responses. In the last two years, there have been a ton of changes in the data center networking space driven by virtualization, data and storage networking consolidation, improvements in hardware efficiency, and new protocols to support increase demand. Typically, data centers are isolated enough from the campus LAN have unique demands that using a different vendor than your campus LAN makes sense. We want to get your input as well on what you think is important.

As you begin to plan your next hardware revision cycle, it's a good time to revisit your vendor and see if they are going to meet your needs in a cost-effective manner. The coming features in data center networking like faster interfaces at 10Gb today with 40 and 100Gb on the near horizon means that switching gear built in the last 5-10 years won't be able to adequately support the speeds or densities in the nest 5-10 years. Take core chassis switches. The backplanes may have the capacity, but the card to back-plane interface will be the choke point. Top-of-rack and end-of-row switches aren't going to be dense enough to support the number of 10Gb interfaces you will be installing starting in the next few years. Oh sure, you will continue to roll out 1Gb interfaces, but you can expect to migrate to 10Gb as the price of interfaces drops, more servers have 10Gb LAN on motherboard, and upstream switches support 10Gb.

New protocols provide opportunities to make more efficient use of your data center providing lossless Ethernet (and who doesn't want that?). Flatter network design makes more efficient use of multiple network paths, and marked improvements in silicon cause significant reductions in power requirements and heat generation.

Network equipment vendors are approaching these challenges in some common and uncommon ways. For example, all the vendors have a way to flatten the network tiers that look pretty similar, but how they address support of legacy equipment, optimize the network paths, and how they are going to support lossless Ethernet vary widely. Equally important, nearly all the vendors have or are developing plans to integrate with SAN, hypervisors management systems, network management, and orchestration systems. There aren't any decent standards work in these areas, so the vendors roadmaps are going to be as important as speeds, feeds, capacities, and cost in the decision process.
 
I'm looking forward to this project. As we progress through the RFI, I'll keep you posted on the progress and whatever interesting bits come to light. If you have any burning questions you'd like asked, just reply in the comment area or drop me an e-mail.