If Heavy Reading’s OSS in the Era of NFV/SDN conference in London yesterday is any barometer, the state of the two virtualisation technologies in telco is, well, as clear as mud. As analyst James Crawshaw put it in his introductory presentation, in ‘Hype Cycle’ terms we are in the ‘Trough of Disillusionment’. There’s not much hype about that.
Here are some takeaways from the event, the first part of a three blog series:
Migrating from legacy infrastructure to NFV is a real conundrum for operators. There might be an appetite to do it, but there’s little common understanding about how. One barrier is the glacial pace of standardisation initiatives: They are necessary, but they simply take sooooo long. As one telco speaker put it, “standards move at a pace that enables you to offer the services today that your customers were demanding eight years ago.”
Did we say pace? That’s also an issue with Open Source initiatives, which telcos need to grasp moves at the speed of the community, not the (business) speed that they may demand. And while we’re talking about Open Source, there’s common agreement that vendors have to embrace it (not least because it’s competitive.)
Still on Opensource the elephant in the room here may be ONAP which may (or may not) force an industry standard based on a whole orchestrator built on Opensource. In oury view to get this to work, together with DevOps and agility, each operator will need to build an R&D department. Whether or not this is broadly practical is open to debate (though either way, it would usher in a golden era for SI’s which in a way contradicts to purpose of DevOps. “Not all telcos are Tier 1’s” was one statement at the event. The question is if whether the options will be OpenSource (tier 1s and advanced tier 2’s) and full stack for the others?
The issue of a Common data model was discussed endlessly, driving the observation that “Pigs will fly out of my ass before we will see a common data model” (we can attribute that to a participant from PCCW.” Either way, there is a need for a common language between domains and operators, which was also addressed, and to have API’s seems to be the widely accepted solution. HOWEVER to handle this a middleware layer will be needed to manage versions, API’s, potential storage (and why not preprocessing too?). One vendor said that Automation needs to rely on a viable information model; probably true, but that doesn’t have to mean it’s standardized.
Another comment on ONAP: It is a bit of an odd animal in the opensource community as it provides such a big framework together with the complexity of involved standards. One observation was that in essence it’s only AT&T and China Mobile that contribute from operator side (unconfirmed) but if true, there are a lot more stakeholders who need to be involved.
Keith Brody & Ola Billinger