— Grant Lenahan, Appledore research Group
There’s a saying:
“You can forward packets, count them or classify them. Pick one.”
For as long as i can remember, there have been trade-offs between telemetry and router performance, and between DPI and router performance. The problem has often been “solved”, if we can call it that, by having tandem systems – one routing and one performing DPI or other tasks on a subset of the stream.
On Wednesday, NOKIA announced its new FP4 packet processor, which claims not only to increase speeds and density by more than twofold, but to do so while delivering DPI, DDOS protection, and rich telemetry, all at once.
So why was ARG, the self-proclaimed cloud management guys, covering a hardware launch and why am I blogging on hardware? Simple – chips such as FP4 will pave the way for more advanced management tasks. We can collect more telemetry data. We can collect that data more often. We can push complex masks to the chip and ask it to pattern match everything. We can do what has heretofore been possible, but uneconomic. We also presume that even if NOKIA is ahead of the industry (as they claim) that others will follow, and we may be seeing the beginning of a new age in analytics and more importantly, in actions.
NOKIA understands that. While about half the day was spent showing-and-telling the “sexy hardware” – (I am not making this up) roughly the other half was spent on what kinds of tasks it could enable. NOKIA showed examples and diagrams that will warm the hearts of any ARG reader – with nested, closed, automated loops operating between the various service routers, Deepfield analytics, and NOKIA’s NSP SDN solution. Many beneficial use cases were discussed, including:
- more capable and complete DDOS blocking that is performed earlier – before it gums up routing
- more insightful global optimization of paths and IP/MPLS traffic using rich telemetry, Deepfield analytics, and NSP
- Optimization based even on external congestion or ingress congestion, e.g.: better peering point selection and allocation (using the usual suspects)
The message is powerful but subtle. From the management perspective, nothing shown was previously impossible – but it was impractical. High telemetry rates and widespread DPI just cost too much, either in money, lost throughput, or both.
What I found most interesting is how much NOKIA found it possible, and desirable to take what was a chip and router launch, and raise the level of discussion to topics of automation. This illustrates what has always been true, but is becoming more evident – the network and its management are two halves of a single whole. The network defines what is possible, the management (and control plane, which are merging) takes advantage of those possibilities. Abstract, expose, consume.
You can find additional summaries of FP4, the new router line, and Deepfield here.
This is certainly the first in a series of similar announcements from leading players in the industry, and promises a future with fewer cost and practical limitations on the data we can collect from the network, and the complex filtering and forwarding rules we can throw its way.