5G's elusive quest for 1ms latency
I have been taking another look at latency in 5G in advance of a talk I’m giving at Understanding Latency 3.0 (https://understandinglatency.com/ ) alongside top experts in network latency, entitled “5G’s elusive quest for 1ms latency”. The conference is 9-11Dec, I’m speaking on the 9th.
As part of my research, I took another look at the implications of an operator offering a URLLC service. I knew that providing low-latency had implications on overall network capacity but was surprised by what I found. For example, one academic paper had the following chart
Without digging too much into the detail this says that for small block size (32 byte) URLLC systems there is a 60x penalty – so 1Mbits/s of URLLC would displace 60Mbits/s of eMBB. For larger block size (200 byte) the penalty is a smaller 10x. There are many other issues associated with signalling channels, but the summary is that URLLC will be expensive for an operator to offer. Logically that would imply a high willingness to pay would be needed.
I’ll explore that and more, including what this all means for 6G in the talk.
Senior Expert at Ericsson Research| Spectrum Management| Radio Research | Trustworthiness | RAN and Physical Layer
3 个月Should the thesis not be that URLLC will pose constraints on the end-user in how accessible it is? In any case, I have always maintained that the drive to reach 1ms link latency was a strawman. There is a value in providing bounded end-to-end latency and many sources of delay in the network stack can add significant latency to information flow. For 1ms round trip latency on the airlink to be relevant there is a lot that needs to happen at higher layers of the stack and in the deployment architecture. All possible. The problem with labeling 5G as committed to low latency or to millimeter wave spectrum alone (both common tropes in media and some evangelical literature) is the simplification of a very complex set of features that provide a systematic set of improvements to 4G. It’s the package we should be looking at. And the standard is a significant improvement over the previous generation. Any discussions on latency should be accompanied by a discussion that looks at use cases, business aspects, customization of solutions for markets that scale up etc. The standard is very capable. The market needs a clear path to adoption and growth. Concepts are free, development less so, and building products depends on a lot of factors.
CEO at RANsemi
3 个月This is why URLLC has always been something that would only be delivered on private networks. In addition, you can't do very much URLLC on a cell or you'll lose any deterministic low latency.
Registered to hear your talk William Webb! Looking forward to it.
CMO at Domos
3 个月Excited to hear your talk, William!