Developer Days 2012 in Silicon Valley is over and I’ve had the time to both travel back home and also let the experience and conversations sink in. And as I hoped in my post before the event, the conversations did span the whole spectrum from enthusiastic demos of our new Web UI to hallway discussions on how we can improve our post-sales process, and to how to better match the varying requirements that development teams have as they progress through their product cycles.
My main takeaway was that management aspects are rapidly moving into center stage after many years of trailing behind other features, most notably hardware and data plane-related. Some of the teams that I spoke to suddenly found themselves being challenged with some profoundly different expectations on their management including:
- Full programmability of all features in the product using NETCONF, REST or other APIs
- High-availability features with real-time or near-real time replication of configuration and operational state across network nodes
- Deep virtualization of hardware and data plane resources to provide resource isolation between “tenants” (e.g. per-customer router instances).
Exciting times, I’d say, but the amount of change this requires on the network vendor development teams is not to be underestimated. It shakes up a whole lot of assumptions on how the system as a whole needs to behave. One telling example of this is that there is very limited support for these types of things in the design and implementation of the commercially available l2/l3 networking stacks on the market. It’s turning into an example of where requirements on manageability start impacting the other parts of a common router or switch architecture instead of the current situation where manageability trails the rest.
In summary, I think we’re seeing the first signs of a reversal of requirement flow internal to networking vendors. Manageability will eventually be the main driver of the product life cycles.