The original concept for edge computing in the telecoms sector stemmed from network virtualisation efforts. CSPs were building distributed data centre-like facilities to support their network function infrastructure and consolidating network functions onto server infrastructure in place of functional hardware appliances. This meant there was spare capacity in those facilities to support customer applications and maximise economies of scale by sharing as much of the infrastructure as possible.
However, reality has diverged from the theory. The requirements for network functions are different to that of IoT/IT/business applications that will be hosted at the edge. Network teams have arguably stricter security rules for controlling data and access and the type of hardware for network functions is different than that needed for enterprise applications. For example, radio access network functions need specialised hardware, such as field programmable gateways (FPGAs) and hardware accelerators. By contrast, video analytics, gaming or AI-intensive applications may need GPUs and other technologies for enhancing graphics processing.
Despite these limitations, there are different elements of the stack that can be shared between network functions and consumer/enterprise applications, beyond the site itself. Our survey demonstrated that there is a wide range of opinions in the telecoms industry on the level of convergence. Only 12% of respondents globally believe that edge infrastructure should be entirely converged (facility, hardware, application platforms and orchestration). On the other side of the spectrum, 13% of respondents feel that the infrastructure should be entirely separate and only the site itself should be shared. However, there are slight differences across regions.
CSP respondents in North America were least supportive of having completely converged infrastructure, whereas Asia Pacific operators prefer convergence. One of the reasons for this is organisational dynamics, which will be explored in the next section. Another reason could be the level of maturity of the operator’s edge strategy; CSPs in North America and Europe have gone through a process of initially attempting to share the infrastructure and operating models, but have since found it to be too challenging to manage each domain’s different needs. As a result, most CSPs in those regions prefer a degree of separation in the hardware and software stacks.
Find more about the findings of our survey and why edge infrastructure will be multi-cloud in our report Building telco edge: Why multi-cloud will dominate