Why the world needs a pervasive Network Compute Fabric
Future generations of networks will bring all physical things even closer to the compute domain. The network will act not only as a connector, but also as a controller of physical systems, ranging from simple terminals to complex and performance-sensitive robot control and augmented reality applications. The network will host computing intertwined with communication for the highest level of efficiency. This is what we call the Network Compute Fabric.
The Network Compute Fabric will rely on ubiquitous compute resources that are fully integrated with the network to complement connectivity, spanning central cloud, network edge, and all the way out to devices. This will allow distributed applications interacting with physical reality to benefit from having their tasks close to data sources and data consumers, for example, in radio beamforming, closed-loop control of mission-critical processes, or intelligent aggregation of large amounts of data. Essentially, the Network Compute Fabric will ensure optimized application performance with real end-to-end guarantees, spanning both the connectivity and compute domains.
Pervasive characteristics of the Network Compute Fabric
In this context, pervasive should be interpreted as “everywhere”. Billions of networked smartphones and connected devices are combined with cloud and edge resources to form an immensely powerful, unified compute infrastructure. This infrastructure has the potential to create completely new, unprecedented applications and services that would’ve been very difficult to build previously.
A non-pervasive compute model would typically be a centralized one. In such a model, huge datacenters take advantage of economies of scale for the heavy lifting, and devices use their limited capabilities for visualization, user interfacing, and sensing. For many applications, this will continue to be sufficient.
However, there are compute and data tasks that need something more. They are too big, complex, or energy consuming to execute on devices, but at the same time it’s not suitable to send them off to faraway datacenters, due to, for example, strict latency or bandwidth requirements, the need for local interaction with other devices, legal requirements, etc. These tasks need options to be offloaded to a closer execution environment. With a pervasive Network Compute Fabric, including a unified execution environment, the application gets a compute service that always appears local, despite dynamic network changes or user/data mobility events.
The network-embedded computational resources will offer unified interfaces, providing access to hardware accelerators and system software infrastructure. This simplifies development of distributed applications and enables seamless task mobility so that processing can follow data sources or users. Taken together, pervasiveness enables uninterrupted and close to optimal application performance, as tasks in the Network Compute Fabric can continuously undergo updated placement decisions in response to changing network conditions or environmental context.
Where trends collide
Networking will naturally be at the very core of the Network Compute Fabric. The emerging 5G networks will be central to getting the right data to the right location at the right time. Taking a step back, we see two major trends colliding to drive the emergence of the Network Compute Fabric.
- Digitalization and cloudification is ongoing in most industry verticals, and so is the trend to push services (applications and their execution infrastructure) closer to the users and data. Current use cases already show the need for edge and distributed computing and storage capabilities embedded into the network to support applications that need certain tasks performed close to data sources or consumers, like AR/VR, automotive, and industry 4.0.
- We are proud to conclude that right now there is no bigger distributed system on this planet than the telecom system. Digitalization is also ongoing in the telcom industry, which means that more and more general-purpose hardware and software components are used to build today’s networks.
Looking at these trends, we can see that this might be the perfect combination. After all, what is better suited to support applications with a vastly distributed infrastructure than the global telecoms network, which is currently re-furnishing itself with general purpose infrastructure anyhow? With the Network Compute Fabric, we work at the intersection of these two trends.
Pervasive Network Compute Fabric: a gamechanger
With large, integrated applications, for example in industrial contexts, there will be many subsystems with varying requirements pertaining to each one. Taking a production plant as an example, there will be subsystems on the factory floor where real-time, or near real-time operation is required. There will be regulations, internally as well as governmental, for how data can move and be stored, and some subsystems will be on-premise whereas others will reside in a public cloud facility.
With a properly designed Network Compute Fabric, we can postpone the decisions of where and how to realize individual subsystems until the time of deployment, or even re-evaluate and change decisions during runtime.
We are aware that this is, and has always been, a challenging field. There have been lots of attempts at addressing the problems. Java was once hailed as allowing you to “write once, run anywhere”, and more recently we’ve seen µ-services and Functions-as-a-Service proposed as the way forward. One problem with Java was that it was PC-centric and left communication and distribution for developers to solve, whereas with later cloud-centric approaches, communication is transparent to the point where developers have no control over where the code is executed.
Looking at it from a telecom point of view, earlier initiatives to create solutions had a very proprietary flavor and did not take the “open” developer community sufficiently into consideration. The Network Compute Fabric vision, instead, embraces the importance of open interfaces, open-source components, and an open marketplace. The Network Compute Fabric should be seen as a complementary solution that seamlessly integrates and plugs into the existing cloud/IT ecosystem.
With widely deployed 5G networks and early steps toward 6G, the reach and performance of networks are developing to a point where they provide a perfect infrastructure for pervasive and unified computing. The deployment and integration of Network Compute Fabric with other ecosystems is now possible also due to the fact that telecom networks these days can be built largely on top of IT hardware and software components and principles, instead of proprietary boxes.
The future potential of a Network Compute Fabric
In the future, there will be even more use-cases that require local compute and storage capabilities. Some that we have identified include:
- Applications using latency-sensitive and high-volume sensory data (as in the internet of senses)
- Larger fleets of connected machines
- Increasing realization of cyber-physical systems with digital twins,
all of them operated and optimized by distributed AI processes powered by telemetry data from both the physical and digital worlds.
It is still too early to predict all the future values of a pervasive Network Compute Fabric. Which of the potential application areas listed above will eventually create the most value, also depends on the business model established on top of the Network Compute Fabric.
And once the possibilities of the system are understood and made accessible to developer communities, previously unimagined use cases might turn out to be the real killer apps!
Network Compute Fabric research
At Ericsson Research, we have spent many years of research and exploration of both distributed systems (see our continuously published results: Calvin, Nefele) and real-time systems, as well as various industrial use-cases. This experience has guided our research into a pervasive Network Compute Fabric, which we are currently exploring.
Explore more
- Read the CTO technology trends 2020
- Watch the full video about the journey beyond 5G.
- Read our whitepaper outlining a research outlook towards 6G.
RELATED CONTENT
Like what you’re reading? Please sign up for email updates on your favorite topics.
Subscribe nowAt the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.