For being the first event of its kind, the Metro Ethernet Forum’s (MEF) GEN14 event was pretty impressive. They attracted over 1000 attendees, with more service provider attendees than network equipment providers. MEF president Nan Chen’s keynote was bold as usual, painting his ambitious vision for a “third network” that would combine the ubiquity of the Internet with the reliability of the Carrier Ethernet 2.0 networks that provide enterprise grade services today but still experience slow provisioning times. Though the rest of the conference focused on more specific and tangible issues, most of it did fall under the umbrella of service orchestration and similar themes, which one could argue is the missing link to that third network vision.
I wasn’t keeping count of the technology terms most used throughout the conference, but I’m pretty confident that these three would be pretty high up on the list:
- Network Function Virtualization (NFV)
- Operations Support Systems (OSS) / Business Support Systems (BSS)
- Software Defined Networking (SDN)
Today Carrier Ethernet services are deployed and seem to work fairly well, but a) the cost models of networking gear when compared with x86 systems and NFV are evolving, indicating potential CapEx savings, and b) it’s becoming more apparent how more intelligent OSS systems, common APIs, NFV, and SDN can streamline the deployment, adjustment and teardown of network services to both save OpEx costs, and increase customer satisfaction by reducing response times. What will this all look like eventually, and how does it look now? Here are a few observations:
1. The next generation OSS / BSS is the Holy Grail, and it won’t be simple.
Enabling a central point at which to orchestrate services was very top of mind. Of course these exist today, but the kind of power that the industry hope to achieve, where virtualized network functions and enough intelligent APIs can provide point and click service deployment, adjustment, and tear-down, which also plugs into billing and customer-facing portals that show performance automatically, is still too complex to orchestrate today. Opinions varied on the different questions: Will providers’ current systems be leveraged or should they start from scratch? Is this years out or do some providers already have this accomplished? However there was little debate on this: service providers will continue to build some pieces of these themselves, but want more holistic and standardized help from solutions providers and vendors – hence the focus on standardizing the systems, protocols, and APIs.
2. The most sought after functions to be virtualized might not be what you’d think.
Questions around SDN were certainly raised, but as expected the event speakers confirmed that NFV is where the current momentum and interest is for service providers. A representative from the European service provider Colt indicated they believe there is an 80% savings to be realized by NFV, based on the cost of today’s x86 architectures when compared to your average networking gear. But what functions will be virtualized? Will core routers be replaced with servers? There was some debate on this with some parties showing skepticism but other functions stuck out as being mentioned much more often, most notably those most associated with service chaining (firewalls, load balancers, etc) and Customer Edge (CE) routers. Services that include a CE being dropped off at each customer site can amount to quite a bit of CEs in total – that can be a pretty high savings multiplier if there are in fact OpEx and/or CapEx savings to be had there.
3. Whether to go multi-vendor or not is no longer even a question. The question is how will everything be integrated.
This was reiterated by nearly every service provider presentation, though there’s still a lot of work to be done on the “how.” Sure providers have multivendor networks today, but the hope is that the increase in APIs and standard tools will streamline the integration of multivendor networks and ease their headaches. A good example of the debate was commentary around the TeleManagement Forum. The TMF has done a lot of great work defining APIs for different use cases and has a lot of momentum, but of course on the flip side some are weary of the number of APIs and number of organizations defining APIs. There was little discussion about transport mechanisms as these, to a large extent, are already interoperable today, and there’s no sign of IP/MPLS going anywhere (though ancillary technologies can still serve a purpose). The conference was an example of how much the conversation has shifted away from transport revolutions like MPLS-TP, and is now more focused on management and the streamlining of provisioning.
One final item of note – My favorite part of the conference actually were the proof of concept demos, which showcased several examples of the kinds of multi-vendor provisioning and OSS type systems that were discussed in the conference in a live (or pseudo-live at least) environment. Here’s a shot (above) of Allstream and Tata Communications showcasing elastic Ethernet service provisioning across both carrier networks, dynamically adjusting bandwidth on the fly and showing resulting video quality.
Are there other perspectives out there? Please do comment or engage on Twitter: @JonathanMorin