There are three types of 5G RAN customers I run into:
This blog is addressed to group #2 to convince them to move to group #3.
The thought process of the group #2 is as follows: “I see standards as a way to get discounts from existing vendors; however, I don’t see enough value in a third-party vendor neutral SMO. The “safety” of one throat-to-choke outweighs the “adventure” of disaggregation, even if disaggregation promises innovations, further cost reductions, or prevents vendor lock-in. So, I’m just going to buy my SMO from the RU/DU/CU vendor.”
Chris Lamb’s presentation from GTC Fall 2022 titled “Using AI Infrastructure in the Cloud for 5G vRAN,'' is fascinating. While it doesn’t directly address the above argument, it can be used to do so. Chris starts by stating that up to 65% of a public 5G network’s CAPEX is spent on RAN. Furthermore, only 25-30% of the RAN is utilized. In other words:
We are talking $10Bs… if not more; this is a huge waste. Clearly, there are a lot of sustainability issues to consider as well.
What if this 46-49% CAPEX could be reduced or eliminated? Chris describes a whole array of AI/ML applications that could be run on the same hardware resources during times of underutilization. In my view, if sustainability was your primary concern, you could simply turn off servers as opposed to using them for something else.
For the above schemes to work, the Service Management and Orchestration component (or SMO + workload/service orchestrator) needs to orchestrate and manage the infrastructure (servers, networks, GPU/DPU), transport, RU, DU, CU, NearRT RIC, and edge computing applications. Given that these components are unlikely to be from one single vendor, the SMO has to be vendor neutral.
I hope you are convinced to move to group#3. You could save your company billions of dollars! And help reduce your carbon footprint as well.