Mobile Networks as a Catalyst for Distributed Computing

Subscribe To Download This Insight

By Dimitris Mavrakis | 1Q 2020 | IN-5734

Distributed computing is a standard Information Technology (IT) industry term that describes computer systems that are distributed across a large area and in many cases positioned adjacent to their end users. In fact, telecommunication networks are large distributed computing platforms, with core networks, aggregation points, and base stations forming a very large computing platform. Nevertheless, in the context of data processing and Artificial Intelligence (AI) model training, the bulk of computational functionality is still very much centralized in the cloud computing model where Web giants like Amazon, Google, and many other companies like Intel, HPE, SAP, and Xilinx are attracting the lion’s share of processing. However, as AI models progress, sensor, training, and inference data will continue to increase and transporting them to the cloud will no longer be an efficient model.

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.

 

Distributed Computing in the 5G Era

NEWS


Distributed computing is a standard Information Technology (IT) industry term that describes computer systems that are distributed across a large area and in many cases positioned adjacent to their end users. In fact, telecommunication networks are large distributed computing platforms, with core networks, aggregation points, and base stations forming a very large computing platform. Nevertheless, in the context of data processing and Artificial Intelligence (AI) model training, the bulk of computational functionality is still very much centralized in the cloud computing model where Web giants like Amazon, Google, and many other companies like Intel, HPE, SAP, and Xilinx are attracting the lion’s share of processing. However, as AI models progress, sensor, training, and inference data will continue to increase and transporting them to the cloud will no longer be an efficient model.

Cellular networks in the 5G era are becoming more than telephone networks, with telco cloud, network automation, and edge computing democratizing the distribution of advanced computing across the cellular domain. 5G is well-positioned to become an enabler for distributed computing and advanced AI models, including distributed and federated learning, but it also faces several challenges before it achieves a critical mass. For example, when private and dedicated cellular networks are being deployed across many enterprise verticals by third parties, telecom networks are no longer the exclusive domain of Communication Service Providers (CSPs). This represents a new set of challenges but also many opportunities.

What Will It Take for 5G to Become a Distributed Computing Platform?

IMPACT


Future cellular networks will consist of public networks (like today’s consumer-focused networks), private cellular networks for enterprises, and hybrid models that will combine components from both fields. Edge computing may be deployed in two ways: as a shared resource and as a dedicated, single customer function. As a shared resource, it is deployed at the edge of the public network (e.g., the cell site), but in the dedicated approach it is deployed on-premises with a specific use for specific applications. For example, an industrial manufacturing company may wish to deploy a private 5G network on the factory floor coupled with an edge computing server to enable machine vision applications for predictive maintenance. In these scenarios, enterprise will not wish to share either the network or the edge server. But for these scenarios to happen, several parts of the ecosystem need to evolve:

  1. Consistent Shared Spectrum Regulations: The Citizens Broadband Radio Service (CBRS) shared spectrum initiative in the United States has shown the world that there can be significant interest in private cellular deployments and the Federal Communications Commission (FCC) approved the full commercial status for CBRS deployments. However, CBRS regulation has been complicated in the United States due to the presence on “incumbent” users in the band: government and military. Contrary to this, the countries that followed (Germany, Japan, and the United Kingdom) have taken different approaches, and in some cases the spectrum application and approval processes are entirely manual. There needs to be a common shared spectrum guideline or policy for governments across the world, so that the supply chain can create economies of scale and multinational enterprises use similar networks across their operations.
  2. Clear Market Segmentation: Currently, there is no clear route to market for edge computing or private cellular networks, where CSPs, vendors, systems integrators, and even the enterprises themselves can deploy a variety of platforms and solutions. This complicates the decision-making process for enterprises and in many cases forces decision makers to choose alternative technologies—e.g., WiFi. Although there is no governing body for this type of deployments, there needs to be a set of guidelines for enterprises to follow.
  3. Acceptance That Private Networks Are Here to Stay: CSPs are largely defensive against private networks for enterprise applications and are pushing for network slicing and public spectrum use. However, several enterprises have very stringent operational data requirements, which necessitate private network and shared spectrum use. CSPs should accept their place in this new value chain and take steps to take advantage of their existing expertise and assets.

These are arguably long-term initiatives, but with the advent of OpenRAN and the entry of Web giants into private cellular through shared spectrum, the industry is at an inflection point which will redefine who becomes the dominant player in the Business-to-Business (B2B) and distributed computing 5G era.

Redefining the Telecoms Value Chain

RECOMMENDATIONS


The telecoms value chain will likely be very different from what it is today. In particular, open networks will allow enterprises and even mobile operators additional choices of network equipment. The ongoing evolution of virtualization will eventually completely commoditize hardware. At the same time, Web giants will enter the market through shared spectrum frameworks and claim a stake in the last mile of connectivity, previously the exclusive domain of telcos. The most important question is how can CSPs remain relevant in the 5G distributed computing domain?

It is now clear that CSPs need to become more open and agile and rely less on their trusted Tier 1 infrastructure vendors for providing all equipment and infrastructure for 5G. Doing so will only restrict them, since this new open network value chain will likely create innovation well beyond the established paradigm. It is also a task for the 3rd Generation Partnership Project (3GPP) and other relevant organizations to understand and embrace the fact that future networks do not necessarily need stringent and copious standards, but rather open interfaces, collaboration between different parties, and reliance on more flexible architectures. This is a transition that will take more than 10 years to fully develop, but CSPs and established vendors need to start planning today, before it is too late. If they don’t do so, this open network value chain—coupled with shared spectrum initiatives—will create a new wave of companies that will eclipse today’s telecommunications giants.