Getting Open RAN acceleration right directly impacts network monetization opportunities
As the Open RAN market continues to mature and evolve, there’s a growing debate around whether in-line or look-aside acceleration is better-suited for lower layer processing.
The main difference between these two approaches is that in look-aside acceleration, only selected functions are sent to an accelerator card and then back to the CPU, while in in-line acceleration parts of or the whole data flow and functions are sent through the accelerator.
Picocom President Peter Claydon weighed in on the debate and tied Picocom’s preference, indeed the preference of many companies involved in the space, for in-line acceleration as it bring positive dynamics to the performance and power consumption of Open RAN systems.
According to Claydon, in-line acceleration is the best option for Layer 1 processing. “There were two camps here really. There’s Intel and basically everyone else, which includes Picocom, Qualcomm, Marvel and NVIDIA. So Intel wants to implement as much of the PHY [Layer 1] as possible on their processors. In fact, they designed processors with specific acceleration in the process of instruction set for Open RAN. They’ve also got a new processor with multiple chips on the same package which they’ve taken the acceleration function and put it in there,” Claydon said during an interview with RCR Wireless News during the RCR Live: Telco Reinvention event which took place in London on April 25.
“Now, Intel’s problem is that they can’t do everything in software and their processors so they want to have this look-aside acceleration model. Other people design the entire PHY on their own chips, so hence, it’s a completely separate thing and none of it is implemented on the x86 processor,” Claydon added.
“Intel starts off saying that because it’s standard hardware it’s going to be cheaper, but what they ignore is the fact that the amount of silicon area you have, it controls the cost and the power generating and silicon basically cost $200 billion per square mile. And so if you have a bigger chip, it’s more expensive and Intel’s chips are enormous and you can implement the same function optimally on something that’s far smaller, as other people do. Hence you’re going to reduce the cost and you’re going to reduce the power dissipation. What happened in the early days of Open RAN is those optimized chips didn’t exist. Now they do, so the equation has changed,” he added.
In terms of the feedback Picocom is hearing from OEMs and their operator customers, Claydon said, “everyone is probably going to move towards in-line acceleration.”
When asked what needs to happen to return operators to growth and profitability, Claydon noted that this process will take a long time, adding that Open RAN will play a role in this process. “I think Open RAN is an ingredient there. I think that’s going to open-up the market to more players from our perspective as a vendor…They’re going to have a mixture of different types of equipment, more of a mix and match, and that’s going to help them increase performance in their network and reduce their costs.”