AT&T's vision for automating network services and infrastructure, unveiled in detail last week, is attracting considerable industry attention. But it's still too early to determine if the massive software project known as ECOMP becomes the unifying force AT&T hopes it can be and is released into open source. (See AT&T Shares ECOMP Vision, Might Share Software.)
In the first week after the announcement, about 3,000 people visited the AT&T page for ECOMP -- which stands for Enhanced Control, Orchestration, Management and Policy -- and about 45% of those have downloaded the white paper on ECOMP which details how AT&T Inc. (NYSE: T) has built this infrastructure delivery platform, says Chris Rice, VP-Advanced Technologies and Architecture at AT&T. Designed to automate the full range of service delivery and assurance tasks in support of virtualized network functions and SDN, ECOMP is intended to present what AT&T executives see as the first complete vision for how to offer a scalable, comprehensive network cloud service.
While there are many other initiatives out there -- probably too many -- Rice notes that AT&T's is different because it has been in use for more than a year in the network, supporting the more than 5% of AT&T's network functions that have been virtualized and the 14 million wireless customers supported on its virtualized mobile core.
"There is real code behind it; it's over 8 million lines of code and still growing. We have regular releases of it," he tells Light Reading. "And besides that, we know how to run networks -- we have been doing that for a while. We've got a pretty good vision associated with this. The best outcome would be that this is one thing that unifies the community around more of a single effort, because there is probably not room for as many as there are out there right now."
He says other initiatives that AT&T has studied fall down on two areas: the ability to scale and the completeness of vision. "And when I say vision, I don't mean that in a soft way, I'm talking about something required to get the job done and to deliver on the promise of SDN and NFV."
That notion of a complete approach to managing and orchestrating a virtualized infrastructure is a constant theme in AT&T's discussion of ECOMP, as is the notion that this is a substantial code base on which to build. Many of the other initiatives focus on the idea of an open source MANO -- the management and network orchestration structure first identified by the ETSI NFV industry specifications group.
"They have shown that the problem space is much broader than the ETSI NFV ISG first defined it," notes Heavy Reading's Caroline Chappell, practice leader in NFV and cloud. "It is not just the MANO, and not just OpenStack, it is the whole virtual network functions automation capability, which is huge, much bigger than previously scoped. From that point of view, it is a very good reference point for the industry."
Rice says AT&T sees what it has done as applicable not just to Tier 1 operators such as AT&T but also to smaller network operators, large enterprises and cloud service providers as well. It builds on the existing layers -- starting at the bottom with a Linux operating system and moving up through the hypervisor layer and the cloud layer to a VNF automation layer.
"The part that is missing, that has to have a broad holistic vision, is what we'll call a VNF automation layer," he says. "And that is required regardless of your size."
Within that layer, again regardless of size, operators want to be able to do things such as orchestration, inventory management, network and application control, data collection and analysis for computing key performance indicators and policy application to enable automation.
"When you start thinking about that, regardless of the size you are, you are going to want to be able to do all of those things, something that synthesizes those capabilities and brings them together in a system in a holistic way," Rice comments.
To do that ECOMP enables essentially a four-layer approach that constitutes AT&T's design and service creation catalog. There are basic capabilities -- think switch, router, etc., -- called resources and putting a series of these together in a pre-defined way creates a service, Rice says.
One key element of ECOMP is that it is built to be VNF- and vendor-neutral, he emphasizes. It is important for service providers to have access to code and the ability to go in and do their own modifications, Rice comments, while allowing things to work together so that "you aren't spending all of your time integrating them as opposed to doing all of the work you and your customers get benefit from."
Building ECOMP to be VNF and vendor-neutral is important, says analyst Chappell, because it may help address the industry's problem with "islands" of NFV that exist today. It's a good idea to have more commonality in what operators want from VNF vendors in terms of automation and manageability, she notes, and to make it easier to onboard VNFs.
"At the moment, we're seeing operators succumbing to buying single-vendor vertical stacks consisting of cloud platform/VNF/VNF management/orchestration on a per-VNF or per-ecosystem of VNFs basis," Chappell says. That essentially means "the industry is back to its old silo approach tricks. This is not going to scale, not going to deliver NFV benefits and perpetuates vendor lock-in."
AT&T's approach keeps faith with the original NFV vision of a horizontal architecture, she notes, with its cloud platform (AT&T Integrated Cloud) and now ECOMP for management and orchestration.
— Carol Wilson, Editor-at-Large, Light Reading