Everything becomes a Cloud

Houston, we have a problem

As CSPs now admit, the plethora of open source initiatives and the disaggregation of the network in fact increased complexity. Quoting from the the AT&T’s Network Transformation white paper, “a significant impediment to NFV and SDN scalability is the current Virtual Network Function (VNF) and Containerized Network Function (CNF) variability across carrier cloud infrastructure”. The challenges in NFV that need to be addressed are:

• Heavy operational overhead integrating VNFs/CNFs with diverse underlying infrastructure
• Increased complexities resulting from varying versions of network function applications
• Minimal VNF provider testing and certification requiring lengthy carrier test cycles
• High carrier development costs due to multiple custom platforms for VNFs/CNFs
• Slow adoption of cloud-native NFV applications

So although the original Software Defined Network, that disrupted the networking landscape 12-years ago, that is itself challenged and almost dead according to some, beaconed towards the right path to open disaggregation, we now have too much openness and the bimodal R&D budget around specialized networking chips (merchant silicon) and microservices (VNFs/CNFs) is endangered by an exploding systems integration cost that is needed to stitch together all disaggregated functions developed independently by various open source projects, together.

The hope of the CSPs that the open source community will take the slack of systems integration has now fully evaporated. One CSP, including opportunity costs, spent something north of $1 billion in R&D of just the software component of now one of the dominant open source platforms in LFN but are now considering withdrawing from this initiative. The reasons are many and go beyond the correctness or not of their network transformation strategy. The strategy is solid in an static environment but stasis is not a word that goes hand in hand with technology companies. The CSPs that defined the strategy in 2012 did not revised and adapt it but like it is customary in standards development, ETSI took 10 years to come up with something that could be the basis of an NFV implementation and that something is 10x more complex that it should be. The market had shifted - millions of enterprises adopted the public cloud and started hosting their most precious assets, their data, in AWS, Azure and GCP. Its not only the data, all Software as a Service applications that power our economy live in the public cloud as well ! So what is left for the CSPs to do in 2020 ?

5G Edge

Quoting from the AT&T’s Network Transformation white paper, “This is part of the evolution towards a distributed, high-performance network edge where customers and application developers can build, integrate, and run highly-optimized applications directly with AT&T’s 5G network while running their application’s on public cloud infrastructure. Additionally, the industry’s growing ecosystem of open disaggregated software and hardware, SDN automation augmented with Machine Learning (ML), Artificial Intelligence (AI), and network fabric automation have already become key facets that are necessary to meet the demands of future traffic growth.”

There is a lot of hyperbolic statements above and the gist you need to notice is highlighted. Its no secret that the automation, the cutting edge DevOps of public cloud providers, cant be beaten easily even with multi-billion R&D budgets simply because CPSs dont have a software DNA and they cant attract the software talent needed. As elegantly quoted “Is Microsoft the new Ericsson?”.

So at&t rushed to partner with Azure, Verizon with AWS

Everything becomes a cloud

The development of open-ran is indicative of where things are going. The up to recently “crown jules” of vendors that accounts to 60% of the total expenditure of the CSP network, are broken up into what is called Open RAN that is effectively a edge cloud with a a bunch of radio interfaces (LTE, 5G).

Public cloud providers via acquisitions are getting into the action and avoiding the dedicated radio functions while focusing their attention on higher value add capabilities such as control (CU), automation obviously via ML/AI. So ultimately everything is a cloud and the job of the formerly called network engineers is dramatically changing as we speak. They are software engineers at heart, with strong DevOps backgrounds, understand the value and know how to extract information out of network data to develop network applications.