AT&T Rallies Carriers Around OpenStack
AT&T hopes to lead creation of a large-scale operator and service provider group within the OpenStack Foundation to bring together like-minded operators and vendors to work together on specific issues regarding the open source cloud platform.
Beginning at the OpenStack Summit in Austin in late April, AT&T Inc. (NYSE: T) is hoping to assemble a group willing to bring resources and developers together to advance OpenStack as a carrier-class cloud platform and do that in an open source fashion, says Sorabh Saxena, AT&T's senior vice president of software development and engineering. (See Showdown at the OpenStack Corral and Showdown at the OpenStack Corral.)
"We want to create a user group like that to take our prioritized needs and work through the foundation as well as project teams, as well as provide our own developers into the community," Saxena tells Light Reading in an interview. Over the next three or four months, AT&T hopes to put together a common charter and determine who is interested and serious enough about the effort to contribute resources.
And that is just one part of a three-pronged strategy on OpenStack engagement that he outlined, to include greater engagement of AT&T developers in core processes such as code review and API normalization, but also working with the OpenStack Foundation itself to develop more centralized processes, in addition to the project-based focus it has today. Calling this "a tipping point" on OpenStack for the network operator community, he says AT&T believes the time is ripe to rally carriers around OpenStack and "put positive momentum forward rather than make it crumble under its challenges."
AT&T is already heavily engaged with OpenStack , as was apparent in Austin, where it won the OpenStack Superuser award, voted on by the developer community, and is using the open source platform as the foundation for its AT&T Integrated Cloud platform. That deployment of 74 AT&T Integrated Cloud (AIC) physical locations in 2015 is expected to hit 105 by the end of this year and reach into the hundreds by 2020, making it what AT&T claims is the largest OpenStack development thus far.
That approach has been described by some as risky but Saxena sees the bigger risks in not being prepared to handle the onslaught of bandwidth demands ahead -- a 10-fold increase in traffic by 2020 -- or in getting locked into a closed vendor solution. And as he points out, AT&T was a founding member of OpenStack and has been at this for a while, earning its share of battle scars in deploying enterprise clouds on OpenStack.
"We are obviously very cognizant and very grounded in the fact we are going into uncharted territory," Saxena comments. "That said, just as a backdrop, the current cost economics just do not allow us the luxury to wait. We switched to open source for various reasons -- the economics, making a community-based approach to get innovation from all sources from the whole cloud community. And very importantly -- getting to the world which we fully understand is an ecosystem, an ecosystem where we need other players but it is an open model as opposed to a black box lock in model."
Traditional processes involving standards committees and RFPs don't enable the access to broad innovation that comes from an open source community or the ability to move quickly, he notes. AT&T's approach involves working with pure-play OpenStack companies and also counting on its own innovation to push OpenStack from a single cloud instance to a scalable platform. To date, those innovations include virtualizing the OpenStack control plane itself, so it could be deployed on a range of sites of varying sizes, without consuming vast amounts of server footprint. "By virtualizing it, we can scale the OpenStack control plane up and down to fit the needs of small, medium-sized and large sites in our distributed architecture," says Saxena.
AT&T then developed an OpenStack resource manager that allows centralized orchestration of that distributed virtualized control plane.
"We built the notion of distributed control plane, meaning every location has to have an OpenStack control plane within it, but you need a centralized orchestrator and a management framework, so we came up with what we call ORM or OpenStack resource manager," Saxena says. He compares it to a hotel reservation system where specific requests/needs are matched to the most appropriate OpenStack location. The ORM was discussed in sessions at the OpenStack Summit and will be "released back into the OpenStack community at the right time," he says.
AT&T also plans to engage with the OpenStack community even more -- it is currently engaged in ten projects but will be adding three more by year's end.
Saxena does see some things that need to be addressed with OpenStack, starting with addressing carrier-grade needs for "supporting performance, latency, throughput, jitter -- all the needs that a carrier has," he says.
Because large-scale operators operate distributed networks -- geographically and otherwise -- OpenStack needs integration with operating tools and a higher degree of possible automation, he adds. And the OpenStack movement itself needs to mature -- the project-based approach that served it well to this point now needs greater integration across those projects to manage things such as upgrades and backward compatibility, he notes. AT&T would like to see the OpenStack Foundation operate more centrally as other open source groups do, and manage things such as testing so individual network operators don't have do their own, he says.
— Carol Wilson, Editor-at-Large, Light Reading