Summary Bullets:
- NFV is (largely) about hosting telecom applications on commercial IT servers. Today, those servers are powered by Intel-based processors.
- ARM-based servers represent an alternative; ARM-based processor vendors are beginning to message the value they can provide vs. incumbent offers.
- Whether or not operators are paying attention is unclear. They need to.
Just about a year ago, I was talking with a telecom infrastructure vendor executive about his NFV strategy when the topic of ‘vendor lock-in’ came up. While software portability promises (among other things) to help operators avoid vendor lock-in, the exec wondered if a reliance on Intel-powered servers meant that the industry was simply trading one form of lock-in for another. We both knew the problems with this sort of argument; after all, diverse vendors leverage Intel processors to build their own kit. No operators will be locked into a single source of servers. Regardless, I played along and asked if this meant that the vendor was developing ARM-based solutions. The answer? “No. Not until my customers begin asking for those types of solutions.”
A year later, Layer 123’s NFV World Congress provided an opportunity to see how vendor (and carrier) thinking had evolved. We’ve already provided our analysis from last week’s event (see: Layer123 NFV World Congress: Trying Hard to Find a Monetization Story for NFV) and this tension between ARM and Intel platforms wasn’t highlighted in that analysis. Why? Because, the thinking seems largely the same as it was a year ago – and that’s a problem.
Problematic or not, this shouldn’t be surprising. Intel-based servers are the industry standard; operators know they work, and that’s important when trying to get a new technology like NFV up and running. We might be seeing ARM-based reference designs for NFV coming out this year, but that’s a long time behind Intel. For its part, Intel’s spending a lot of money on NFV, both in terms of marketing and R&D. The result is solid mindshare and innovation. None of this, however, obviates the need to look at alternatives to Intel-based platforms. In particular, ARM-based platforms promise a number of potential benefits.
- Power Efficiency. ARM is known for power-efficient architectures – important for battery-constrained mobile devices, but also when looking to save OpEx on massive (or distributed) data centers.
- Distributed Architectures. In an effort to differentiate itself in the NFV space, ARM is proposing distributed architectures (supporting diverse combinations of storage, networking and compute) for addressing diverse NFV use cases and deployment scenarios. Could you support these with Intel-based platforms? Probably. Will they work better with ARM? Maybe.
- Interoperability. ARM (and its partners) have been working closely with the OPNFV to ensure that – no matter which partner’s ARM-based processors are powering a platform – an operator’s VNFs work fine.
- Competition. If competition drives everyone to perform at their best, commercial ARM-based NFV infrastructure should help to move the industry forward faster. You can think about competition, here, on two axes. There’s the competition between ARM and Intel-based designs. But, there’s also competition between ARM-based suppliers in ways that aren’t matched in an Intel-dominated ecosystem.
Yet, for all arguments in favor of ARM-based NFV infrastructure, there’s a rub. ARM seems committed to the space. It should be; there’s a solid opportunity here and ARM has a good story. But, nobody can say how long it will remain committed. If operators don’t begin showing their interest, ARM’s commitment will logically wane; it is a business, after all, and needs to invest its time and resources where it can expect a return. And with ARM-based solutions also expected to power equally hyped spaces such as IoT, it’s not as if the company doesn’t have ‘other fish to fry.’