Most network builders and operators are now making transition plans from 4G/LTE networks into 5G. Transmission speeds used to be the main driver for painstaking upgrades from 2G to 4G, but this time it’ll be different, as we may finally be seeing a fully IP-embraced environment without the ghosts of ISDN, X.25 or ATM.
Networks from 2G to 4G were mostly implemented with purpose built hardware and physical servers providing key services like subscriber registers and message queuing and delivery. This way introduced many inefficiencies and wasted computing capacity, as the service infrastructure was rigid and did not support dynamic load balancing. Moreover, network infrastructure providers continuously solved performance problems by just throwing more hardware at software problems, as the finances then allowed for it.
In 5G, which has been on the drawing board for quite some time already, most network related computing functions are destined to be virtualized into a common, off-the-shelf infrastructure, meaning that new kind of providers like Dell EMC and Oracle VMware will have a lot bigger share of any new network rollout. Most vendors call this Network Function Virtualization (NFV).
Having a fault tolerant, standardized backend for all network services is an adaptation from the enterprise world, where virtualization has eliminated most underutilized physical services during the last 15 years. However, although the goal of virtualization is to simplify the architecture, it actually requires more planning as shared services easily affect each other if bad architectural choices are blindly replicated to the virtual world.
Now what comes to network operators’ new services to their customers, having a powerful NFV and edge computing environment presents interesting opportunities to take on the public, everyman’s ‘serial virtualizers’ like Amazon, Google or Microsoft, especially on the enterprise side requiring more fault tolerance and better service assurance. This could be good business for 10+ years from now.
While we welcome NFV as an intermediate building block to streamline network operator’s architecture, it must be realized that the (Linux) virtual server should not remain as the basic unit of computing for too long, as it is a historical construct and still very much a compromise in most aspects, dating back to the 1960s. Containers and microservices are a step to the right direction, but ultimately we’ll see pure programmatic spime containers without the traces of monolithic kernels and other legacy adaptations. Embedded finally meets the rest of the computing world.