Volta has written a white paper on edge computing, has a web page on it, and did a webinar with Jennifer Clarke who covers cloud infrastructure & edge computing for Omdia/Heavy Reading. Why is a routing vendor so focused on edge computing?
Well, the answer is in the name. It is clear that edge computing is a major driver for the new service edge. Since we live in an IP world, we think of the service edge as the location where layer 3 services reside. Using MPLS terminology, an L3VPN is managed from a Provider Edge (PE) router. Those routers are typically large legacy routers housed at a PoP in a major metro. This works fine as long as the services actually make it that far but that is changing, and edge computing is critical to that.
There is a broad range of business drivers for edge computing. They boil down to two main technical issues. First is bandwidth. IoT enables literally billions of sensors to collect and transmit data. Individually, each may not consume much bandwidth. In fact, 5G introduces NB-IoT to support low bandwidth devices. However, the cumulative bandwidth becomes an issue because of the sheer volume of devices. Thus, it makes economic sense to aggregate and process this information at the edge rather than use the bandwidth necessary to move it all to remote cloud data centers.
The second consideration is latency. Applications like process automation, connected cars, and smart cities can all require very low latency that cannot be delivered by distant cloud data centers. Having the data processing at the edge solves that problem.
Latency is also an issue for RAN disaggregation. In our 5G transport white paper, we discuss the options in disaggregation and how they affect the transport network. For example, operators and move the DU and CU away from the cell towers. This allows for savings from economies of scale. It also allows these functions to be virtualized and run on x86 hardware. There are latency requirements that the transport network must meet. In the webinar, we discussed the synergies between the need for edge computing to support both customer application as well as internal NFV.
Thus, there are compelling reasons why the service edge must move from the PoP and be much closer to the customer. This means that the layer 3 service intelligence must be located there as well. It is hard to manage QoS, traffic engineering, and MPLS without a router. That is why TIP’s OOPT has developed the specifications for the Disaggregated Cell Site Gateway. It would connect a small number of cell sites as part of the 5G transport network. In addition, the DCSG can enable network convergence by connecting broadband and enterprise services on the same transport network.
This new service edge must be able to support a broad range of services. Starting with RAN disaggregation, this will expand to encompass everything from low latency (URLLC) to high bandwidth (eMBB). Network slicing will be essential to delivering these new services. This requires the ability to spin up virtualized resources quickly. Since the service edge is now close to the customer, this needs to extend to the router so that separate virtual routers can be used right at the new service edge.