Featured Story
How Huawei went from Chinese startup to global 5G power
A new book by the Washington Post's Eva Dou is a comprehensive and readable account of Huawei's rapid rise on the world's telecom stage.
Dish is replacing VMware with Wind River, which says it beat hyperscalers to the deal and will provide a 'core to edge' platform.
Broadcom's ownership of VMware seems to be working out very well – for Wind River, one of VMware's main rivals in the telecom sector. Commercial sensitivities prevent Paul Miller, Wind River's chief technology officer, from identifying VMware as the vendor it is booting out of Boost Mobile, the brand name used by Dish Network. But he will confirm that Wind River is replacing the "incumbent" provider of container-as-a-service (CaaS) technology. And Dish had previously identified VMware as that provider when listing its army of suppliers.
CaaS is best thought of as the management and hosting platform for the virtualized network functions of a telco. The suspicion is that EchoStar, Dish's parent, was unhappy about price rises VMware has foisted on customers since Broadcom's takeover a year ago. "People aren't holding an architectural gun to my head," EchoStar chief technology officer Eben Albertyn told Light Reading last month, revealing before the Wind River announcement that Boost would be switching to an alternative vendor.
But Miller insists technology and not money was the deciding factor. Wind River has clearly proven its capabilities in the commercial network of Verizon, where it is supporting the world's largest deployment of a virtual radio access network (RAN). Its big claim is that rivals have struggled to be as efficient at the "far edge," an umbrella term for deployments at mast sites and other locations near end-user devices.
Wind River has been able to squeeze overhead software into a single processor core, says Miller, while his competitors need anything between four and 12. In a commercial network, that could be the difference between maintaining just one server at a site or needing two. "When you're times 10,000 sites, it has a horrible impact on TCO [total cost of ownership]," he said.
Upstaging Outposts
Besides representing a loss for VMware, success for Wind River at Boost also looks like bad news for AWS. The US hyperscaler underpins the Boost rollout, which relies on AWS "local zones" and infrastructure to host much of the 5G network. VMware was previously described as an "abstraction layer" in this setup, giving Dish the freedom – if it so wanted – to migrate easily from AWS to another public or private cloud environment. But AWS has its own CaaS offering and would undoubtedly prefer not to have another company sitting between it and Boost for some functions.
What's clear is that Wind River faced competition from hyperscalers to be the VMware replacement. "We did see them, but they were unable to capture the business," said Miller, describing competition for the work as "pretty significant." And while he declines to call out AWS, he is critical of Outposts and Wavelength, the AWS offerings for far-edge rollouts.
"When you look at things like Outposts and Wavelength, far edge to them is just stepping out of their public cloud infrastructure and getting into a carrier network," said Miller. "That's actually four layers away from what is actual far edge." He groups the AWS technologies with the competitors needing multiple cores and more hardware when StarlingX, Wind River's platform, can do it on a single core.
If there has been criticism of Wind River, it is typically that it can only do far edge. Anyone needing CaaS for core network as opposed to virtual RAN functions is better off with a hyperscaler platform or IBM-owned Red Hat, say critics. But the contract with Boost could silence Wind River's opponents. It covers the core network functions as well as virtual RAN, says Miller, and will ultimately give the telco "a single technology stack, core to edge, on which to load all the applications."
Taking edge expertise and repurposing that for the core is more straightforward than attempting the opposite move, he insists. "You have high availability, geo-distributed accelerator support, real-time applications, and so it's a very intense challenge to build a product that works well with that far edge and the core actually now becomes kind of easy," said Miller. "You don't have any of those challenging problems like hardware accelerators and real-time applications."
Within Boost, an extension of Wind River from virtual RAN into the core would conceivably threaten AWS, which has cited the use of orchestration technologies such as EC2 in its blogs about the deal with Dish. The question for the operator is whether the efficiency benefits of using one platform would outweigh concern about such heavy reliance on a single vendor.
A few short months
Its willingness to lean heavily on Wind River, then, could be determined by the vendor's ability to switch out that VMware abstraction layer without any service disruption. Miller expects the whole job to take "a few short months" and says it will not necessitate any hardware changes. "We'll be displacing the current CaaS technology provider with our technology on the same deployed hardware," he said.
This would appear to be Dell servers equipped with Intel's central processing units (CPUs), with Miller conceding there is today a lack of commercial alternatives to Intel in the virtual RAN market. The main hardware challenge would seem to come when the supplier of virtual RAN software has used an accelerator to handle some functions.
"We in-built the management of the accelerators so the cloud actually understands that there's an accelerator present, and it manages the drivers and software associated with that accelerator," said Miller. "We treat the accelerator as part of the cloud infrastructure, rather than deploying a cloud and trying to add drivers for an accelerator separately." The "incumbent" CaaS vendor – meaning VMware – was not able to do this, he said.
Miller similarly downplays the effort involved in the CaaS transition for Boost's virtual RAN software vendors, which appear to be Mavenir and Samsung. "We already have existing strategic partnership alliances signed with them," he said. "There will be integration tests for different things that are unique to the Boost network, especially with the automation layer, but largely we don't see any problems there."
It's a satisfying end to the year for Wind River, which has been owned since 2022 by software company Aptiv. Telcos enthusiastic about virtualization and the cloud will be watching closely to see how well Miller's team handles that migration off VMware – and if Wind River measures up in the core as well as it claims.
You May Also Like