Network Compute Fabric
Network Compute Fabric is a global system of interconnected components such as cloud computing, storage, networking, wireless connectivity, and mobile devices. It supports a rich and flourishing application ecosystem and has become all-pervasive.
The emergence of the Network Compute Fabric was accelerated by the convergence of two major trends– faster, more reliable connectivity and widespread availability of computational services driven by virtualization. It is rapidly evolving toward AI-enhanced cyber-physical systems to enable future applications.
What is the network compute fabric?
Exposure is a key component of the Network Compute Fabric, ensuring that services and capabilities are at the disposal of developers in an easy-to-consume fashion, preferably through APIs.
To support the fast time-to-market goals of modern software development methodologies, the network will adapt to provide quick, easy, low-cost, and resilient service and data exposure creation.
Many connected mobile devices lack either the capabilities or the energy capacity to run certain applications with high quality-of-experience for an extended period. A compelling mitigation to this is to offload the critical parts of the application to a location with high computation capacities and a steady power supply.
A compute offloading service exposed and bundled with mobile connectivity services offers both application developers and users a convenient and easy way to extend device capabilities whenever needed.
The Network Compute Fabric also enables extensive in-network computation. Modern transport networking equipment will no longer be limited to packet transport, but also provide programmable compute capabilities. This will pave the way for new areas of application functionality where data manipulation tasks, such as IoT data process or media encoding/decoding, can be performed directly in the data plane, as part of the communication pipeline.
Deterministic application performance
Current and emerging use cases often require deterministic performance, which includes characteristics like low latency, high throughput, high reliability, and scalability. To meet end-to-end performance needs, it is therefore necessary to supplement the deterministic and reliable connectivity offerings with corresponding compute capabilities. Connectivity with an integrated real-time compute stack will enable the network platform of the Network Compute Fabric to host and manage critical real-time applications, ensuring end-to-end application performance.