For the most part, my last post was concerned about what Cisco ONE was, so explore a little more into the why. I am going to assume you read my last post, so let’s dig in. One of the fundamental concepts behind ONE is illustrated below--the idea of exposing the network in a highly granular way and emphasizing the ability to not only exert programmatic control over switch behavior, but the ability of the network to present interesting and useful information back up to the applications.
This gives us a foundation to move beyond the current ships-in-the-night interaction between apps and infrastructure that exists in most environments today. A simple example I use is firing up a video chat client to call a friend. At the end of the day, the chat client has no idea what actual network conditions are like and simply makes some assumptions about bandwidth, latency, etc. The result can be a wide variation of call quality for call-to-call or even within the same call. Ideally, if the chat client and the network could share information in real-time, it would lead to a much more consistent experience. With Cisco ONE, the goal is to be able to establish these kinds of feedback loops.
They can be very basic of they can be very sophisticated, but the concept is the same: to be able to tap the intrinsic intelligence in the network, assess it, and take action on it. It is a simple yet powerful concept. Take a listen to Igor Gashinsky of Yahoo! talk about self-healing fabrics at this year’s Open Networking Summit. The whole talk is good, but the salient part starts at the ~ 11:40 mark (note: this should not be taken as an endorsement of Cisco or ONE):
As you can see, pretty cool stuff in terms of lowering costs, simplifying management and preserving the customer experience. You could see the same concept applied to security for example, where the analytics are now looking for anomalous behavior. That being said, while I expect a lot of the initial use cases for Cisco ONE to be around simplifying network mgmt, if thats all we get out of network programmability, then I think we have missed the boat a bit.
One of the demos we had on the show floor at CiscoLive this month was something we called “Routing for Dollars” and it shows how traffic could be dynamically re-routed based on changing costs (this is not admin cost, but actual cost). The idea is that using onePK customers can write apps or tools that can incorporate both business logic into your traditional network policy controls. Here is the onePK team talking about this a bit (~3:55 mark):
So that business logic could be all sorts of interesting things like ITIL policies or your governance, risk and compliance (GRC) control systems. I think if these types of solutions start to emerge, things get very interesting very quickly. We are not going to get there tomorrow--I still think some of the simpler network operations solutions will emerge first--but the potential is certainly enticing. We certainly believe we are doing our part by offering up a rich, flexible development kit to facilitate the bridge building with onePK.
Apparently I am in the mood to link to video tonight, so let me close with one final video of a keynote I saw today at the Structure Conference. The keynote was by Lew Mooreman, president of Rackspace and its worthwhile watching the whole thing, since Lew also talks about the value of programmatic access for entire cloud stacks--Rackspace is a huge advocate of OpenStack, which is something that Cisco is also very much involved with. Anyway, if you skip to the 2:30 mark, you’ll see a great metaphor for what we are trying to do with Cisco ONE and onePK.