As you might have noticed, everyone is talking about Data Centers lately and all the new “revolutionary” networking technologies are targeted at this segment. The reason is simple: server virtualization (not to mention the vapor-word ) will forever change the networking landscape and the networking engineers might get badly hurt if caught unprepared.
Traditional data center had a networking infrastructure (“us”) connecting servers (“them”) to the rest of the world. The servers occasionally had some weird connections to external disks (called SAN to differentiate it from LAN) that we tried to avoid as much as possible (after all, they were the responsibility of the storage team). The focus of networking and server engineers was widely different; Greg Ferro published one the best illustrations on this conceptual gap .
Server virtualization and LAN/SAN convergence changed all that and the talk has already shifted from what and why to how . All of a sudden, physical servers contain virtual bridges (oops, I have to say switches these days) and network and storage traffic are converging on the same Gigabit Ethernet infrastructure. The situation is quite similar to the early days of VoIP deployments, with a few significant differences:
- This time, the server teams are the “heroes”. Server virtualization is where most of the savings will be. You will just have to buy even more expensive boxes to support FCoE requirements.
- With converged LAN/SAN/server landscape, a few people will become redundant. In most IT organizations the server teams are larger than the networking teams. Guess what’s likely to happen and which team will be merged into another one.
- Many server engineers have traditionally viewed networks as an obstacle between them and the end users. Simplifying the network by (for example) reducing the whole data center to a single transparent L2 domain with routing at layer 2 would be dreams come true. Why do you think so many vendors tout the advantages of TRILL or its fabric equivalents?
The networking might still be the most important IT infrastructure , but that fact will not help you (or anyone else) when a huge data center with layer-2 protocol-independent brouting melts down (even an “I’ve told you” statement won’t be of much use).
Regardless of what your relationship with the server team is at the moment, you’re one of their best assets (although they might not know that yet). Deploying converged LAN/SAN infrastructure and designing the right mix of L2/L3 switching that will survive the unexpected failures is just the right job for you. However, to make yourself truly useful, you have to grasp the big picture, understand the impact (and relevance) of the emerging DC technologies, start speaking their lingo and start working with them (not provisioning switch ports for them) to help them solve their problems.
A lot of you are already there (and I know I’m preaching to a very large choir); if you’re still missing an few bits, you might consider registering for my Data Center 3.0 for Networking Engineers webinar .