I’ve initially joined Juniper Networks in 2001 and over the years i’ve had the opportunity to establish a relationship with a few of the field people, specially in Europe where i just happen to know a lot of the old timers that built up Juniper’s business in the region.
Over the past few weeks i’d a couple of conversations with some of them that forced me to try to distill my perspective on the current trends in the networking industry to a small set of observations. Often the question that starts the conversation is how I see the applicability of OpenStack and OpenContrail to the key networking markets: carrier, enterprise and cloud/content provider. The question often implies a certain doze of healthy skepticism.
OpenStack and OpenContrail are tools; the evolution that we are seeing at that moment in the industry is deeper than that.
The traditional workflow for a network deployment is to go through architecture, design and operations phases. Traditionally the architecture group selects the top level goals and the technology approach for the deployment and produces an architecture document; from that document the design team then starts working on qualification of equipment, detailed design and operational guide; when a network goes into production stage there is typically an operational guide produced by the design team that is often 100s of pages long that describes the design details as well as operational procedures; operations is then in charge of following this guide: monitoring alerts and implementing the procedure specified in the ops guide. The role of networking vendors is more than anything else to help the potential customer through architecture and design stages: the actual equipment is often the least interesting part of the relationship between operator and vendor.
This traditional model of network design and operations is being disrupted; network operations is moving from humans implementing operational guides to computers; and a smaller number of humans developing monitoring software.
The output of design needs to be as prescriptive as before but using software programs rather than unstructured text; architecture needs to select technologies that are above all automatable and drive the automation process.
Is this is a short transition in terms of time scales ? Absolutely not. Traditional PTTs for instance have a work force that is huge and for which the cultural change will be very difficult. But there is still evolution happening; and there is a natural renewal in the work force: PTTs have a non-trivial number of people that retire each year, for instance.
Automating network operations is something that touches the core of the business model of a network operator and that can significantly transform the business. It is not a temporary fad; this genie is not going to back into the bottle. And the transformation affects not only the operator; the network vendor now needs to bring a very different set of skills and solutions to the table.
Coming back to tools: OpenStack is going to be transformational for many network operators. It is going to be used to drive a new set of operational principles; beyond it direct applications. There is no carrier today that isn’t taking OpenStack seriously.
In enterprise, the biggest change vector is SaaS. Enterprise is migrating to a SaaS model for anything but home-grown applications; and that means that SaaS vendors become large scale data-center and network operators. With the advantage that they get to decide how to build operations after Amazon, Google and the rest of the hip crowd showed the way.
Usually I manage to get a free dinner out of this story, so i’m sticking to it.