The inimitable Steve Kaplan just made an excellent point about Cisco’s new UCS platform over on his blog: calling UCS a “server” is missing the point.  The Hoff has also talked up UCS (he even accuses himself of being a Cisco Fan Boy!) when he said Cisco wasn’t getting into the server business.

Your outfit defines who you are at the office party

Your outfit defines who you are at the office party

The flippant part of my brain compares this to going to rave in a suit, a school disco in beach wear, or a Glastonbury festival wearing Versace: the other people in the party just won’t “get it”, no matter how fantastic your outfit.

If you swap the party for an enterprise IT organization, where the cliques equate to architectural and operational silos, then imagine UCS walks in – what is the response? Everyone knows the Cisco name, but what on earth is she wearing?

Cisco is a network company, right? And surely Cisco only understands layer 4 downwards, right? So in one corner of the party the network guys are sipping their Guinness and are shocked to find that UCS isn’t wearing skimpy, network only, beach wear – no, this outfit is much more complete because it reaches out to storage and compute, and guess what: it’s easier to operate too. A fully clothed platform? Does this make sense? The one flamboyant member of the network team realises that UCS makes the network the centre of the world (wasn’t it always?) and rushes to shake hands with the newcomer.

On the other hand, the Security guys are eating their pitta-and-humous dips with eyes slitted with suspicion. Fibre channel over Ethernet, eh? So you want to send multiple VLANs, multiple protocols, all out of just a couple of NICs? So my un-encrypted fibre channel data goes over the same NIC as the DMZ? Suspicion. Doubt. Resistance.

The Network Operations Center: The Window to Data Center 3.0 / UCS

The Network Operations Center: The Window to Data Center 3.0 / UCS

The operations team seem to be the ones that make most sense of UCS. They are the ones on the front line, managing the seven thousand x86-based servers in three datacenters, not to mention all the other legacy stuff like Novell.

Operations has the dual and conflicting pressures of improving their service whilst reducing cost. They have one hand tied behind their back in the form of legacy design constraints, such as “air gaps in the DMZ” and “design capacity for possible peaks, not actual usage”. They feel the cost pressures in the form of reduced man power, and spend most of them time on mundane activities in a “don’t touch anything you’ll break it” fearful environment.

When Operations looks at the UCS they see an answer to their agility vs. cost dilemma:

  1. These simple “wire once walk away” chassis can handle hundreds of virtual machines each, which means most work can be done from a console (work from home?) and the assets can be continuously redeployed to meet demand (e.g. one month development project only needs ten servers for thirty days: try doing that with physical servers). One UCS chassis can deploy 150 VMs in eight hours – now that’s what I call high performance operations (thanks to Steve Kaplan for pointing this out!).
  2. Now we can provide something like a cloud on top of UCS and move away from static physical servers that are over specified (2 CPU, 2 GIG etc) and under-utilized (~ 6% CPU). With the UCS architecture we can have a higher density. With the UCS management approach we can codify our procedures in software and orchestrate at a much lower cost and higher reliability = maturity.

The Operations guys at the party understand this and are celebrating with tequila shots: they see that UCS is wearing the same clothes that they are, but they have been upgraded somehow.

UCS makes most sense to those on the business end of IT, including those application developers and owners who will be on the receiving end. It will be of great interest to those who’s livelihood depends on high performance IT operations, like hosting organizations who’s business model depends on it, and outsourcers who keep the contract by improving service and reducing cost.

Operations don’t see UCS as a server, they see it as an evolutionary step: an enabling technology that stands on the shoulders of giants (vSphere, FCoE, etc) – just like Steve Kaplan says: Kindle is not just another media reading device, and UCS aint just another server!

Next challenge for Steve Kaplan: a TCO/ROI/Payback model for UCS :-)

Related posts:

  1. Unified Computing = Unified People, Process and Technology