Amazon Web Services (AWS) is courting telcos at this week’s MWC 2021 event, explaining all the services that it can offer to service providers.
Ishwar Parulkar, chief technologist for telecommunications at AWS, went into some detail about the cloud provider’s virtualization stack, its in-house silicon and its edge computing chops.
He said network functions virtualization (NFV) for telco data centers requires a virtualization stack that is very high performing. “NFV needs a virtualization system that is on par with hardware routers that can do very high-speed packet processing,” said Parulkar. “We at AWS build our service based on the Nitro stack.”
The company has a dedicated Nitro chip that handles security and monitoring of the hardware. It also has a dedicated Nitro card that offloads several of the software functions of the hypervisor onto a dedicated card, creating a lightweight, high-performing virtualization layer. “With the Nitro security chip and the Nitro card, we can now have a very lightweight hypervisor responsible for CPU and memory allocation giving very close to bare metal performance,” said Parulkar.
The telecom exec said several of AWS’s innovations derive from the fact that it builds its own silicon in-house.
“We have developed chips that run the Nitro system,” said Parulkar. “What doing custom silicon gives us is a very low cost, low power, low footprint incarnation of hardware that is especially useful in constrained edge sites that telcos have.”
He also cited AWS’s Gravitron2 processors for cloud workloads, which AWS developed in-house, as well.
“Now, if you consider telco workloads like the radio stack, which requires very high performance and needs to reside on edge infrastructure, which is in very constrained locations like cell sites where you have constraints on power, where you have constraints on the real estate, you need silicon that is very efficient in terms of these parameters. And that’s what Graviton2 gives you,” he said.
AWS and the edge
In terms of edge computing, he said the company takes a holistic approach to create consistent infrastructure from all kinds of geographical locations at the edge, to the internet, and to the cloud. AWS starts from its AWS regions.
Incoming AWS CEO Adam Selipsky said yesterday that AWS availability zones can be thought of “roughly as data centers,” and multiple availability zones constitute an AWS region.
Parulkar said that starting with the regions where most of the workloads run, the company extends its infrastructure to various degrees to the edge. He cited AWS Wavelength, which integrates small data centers into telcos’ 5G networks. And he also mentioned AWS Outposts, which are AWS infrastructure that telcos can install in their on-premises data centers, but which are managed by AWS.
“In building this edge infrastructure, or this continuum from the region to the far edge, we have followed one fundamental design tenet and that is consistency of experience and operation,” said Parulkar. “So the AWS infrastructure in all of these sites is the same infrastructure that we run in the regions.”
Finally, he said AWS has tried to optimize the specific SKUs that telcos might need because of either their workloads or “constrained” edge locations such as cell sites. In addition to its 42 unit racks, it also offers 1U and 2RU servers that run on Intel and Graviton2 chips.