Home Search Center Intelligent Model Selection IP Encyclopedia

What Is Computing-Network Integration?

Computing-network integration bridges central and edge computing facilities over agile, reliable, intelligent, and secure network connections, enabling the unified orchestration and control of multi-layer computing resources and the fast, secure, and intelligent transmission of computing power on the network.

Multi-Layer Convergence Empowered by Computing-Network Integration

Computing-network integration literally means to integrate computing and network resources. This concept is quite similar to the Computing First Network (CFN). In fact, a CFN is a transmission network composed of computing and network facilities, and its ultimate goal is computing-network integration. In terms of architecture, computing-network integration proposes a collaborative system featuring four layers and six convergences.

Four Layers

  • Infrastructure: Computing and network facilities are used to perceive, connect, and coordinate various forms of computing power before finally integrating network and computing resources.
  • Platform: The intelligent O&M platform integrates services and computing resources for fast computing power scheduling.
  • Application: Ubiquitous computing connections are provided for various vertical industries.
  • Security: Endogenous security architecture design ensures the secure and reliable scheduling of computing power.

Six Convergences

  • Service convergence: The O&M platform features cloud-computing-network-security integration. It guarantees the quality of computing services, converts computing resources into water- and electricity-like service resources, and reports quality issues, computing resource distribution, and other information in real time.
  • Capability convergence: The computing-network platform integrates various capabilities, such as intent awareness, elastic service, and fault monitoring, to respond to diversified computing requirements in real time.
  • O&M convergence: Network, computing, and storage resources are stored in the same resource pool and combined with deterministic, intelligent O&M, and computing technologies to build a converged and intelligent O&M system that features fast integration, unified orchestration, and unified O&M.
  • Data convergence: Collection, configuration, security, log, and other data is stored in a data pool to form a data middle-end, which ensures data security and agile scheduling through big data learning and intelligent analysis.
  • Computing power convergence: Computing power management, application, and visualization capabilities are provided for heterogeneous computing entities such as CPUs and GPUs. Technologies such as computing power allocation ensure the flexible application of ubiquitous computing power, meeting the computing requirements of various services.
  • Network convergence: Cloud-network-edge-device integration ensures comprehensive computing-network integration.

Why Is Computing-Network Integration Necessary?

New digital infrastructure underpins the overall upgrade of the real economy. In the digital economy, smart scenarios such as biometric recognition, smart healthcare, and smart manufacturing continue to emerge. The digitalization of increasingly more personal and social information leads to massive application data, and computing requirements increase accordingly. It is estimated that by 2030, the computing power demand of unmanned driving, blockchain, IoT, and AR/VR alone will be 300 times higher than it was in 2018. High-real-time and efficient computing is about to become a major characteristic of intelligent scenarios. Efficient computing has been recognized as a priority by major countries across the world. In 2020, the scale of global computing power reached 429 exa-FLOPS (EFLOPs), which was a year-on-year growth of 39%. China was neck-to-neck with the US, Europe, and Japan in computing power scale. It is safe to say that computing power has become a resource that is more valuable than the traffic itself.

However, the construction of computing facilities is currently facing challenges such as high energy consumption in data centers, low computing resource utilization, and uncoordinated and unbalanced regional development. With the acceleration of digitalization and intelligence, the imbalance between supply and demand of computing power is now more prominent than ever before. The reason for this situation is two-pronged. First, the computing industry has been focused on making single inroads into computing power over the years, paying little attention to computing power connectivity. End-to-end computing services lack easy access. Second, the computing industry is still coping with the sharp increase of computing requirements simply by expanding the capacity of the computing power transmission network. This measure is unable to address ubiquitous computing requirements. As a result, local computing resource surplus and global computing resource insufficiency coexist.

The adaptation of networks to globally schedule and connect computing power accelerates the evolution of networks towards computing and intelligence, facilitating the innovation of computing power connection services. Therefore, achieving network-based computing has become an important direction for the communications industry.

What Are the Characteristics of Computing-Network Integration?

1. Flexible Connection of Computing Power

Computing power will be ubiquitously distributed in the future, requiring CFNs to provide agile access. And SRv6 plays a pivotal role in fast access to computing power. SRv6 enables the automatic establishment of E2E paths across multiple levels of networks and the fast transmission of computing power along low-latency paths. In addition, it provides one-hop-to-cloud and one-connection-to-multiple-clouds for you to quickly access computing power and cloud-edge-device resources.

2. Lossless Transmission of Computing Power

When computing power is connected through the network, packet loss may occur when a network fault occurs. Each packet loss has a great impact on computing efficiency. It is therefore essential to guarantee lossless transmission of computing power over the network. Network slicing provides dedicated channels to ensure the lossless transmission of computing power. It offers the following assurances:

  • Lossless Ethernet is adopted by the hyper-converged data center network to achieve zero packet loss.
  • Slice tenants are securely isolated from each other, and device-edge-cloud computing power is reliably transmitted and arrives on time.
  • The intelligent algorithm ensures zero packet loss at 100% throughput and 100% release of computing power.

3. Intelligent computing-network scheduling

When a requesting party initiates a computing power request, the computing-network brain determines whether a network connection service can be provided between the computing-network provider and the requesting party based on factors such as network bandwidth, delay, jitter, and reliability. If such a service can be provided, the computing-network brain uses a multi-factor network measurement algorithm to compute network paths in real time based on SLA, bandwidth, and other information. It orchestrates computing and network resources in a unified manner, observes computing power and network status in real time, and computes optimal transmission paths based on information such as computing requirement changes, delay, and bandwidth.

4. Computing resource awareness

The CFN has massive numbers of application connections, and the computing resource awareness system flexibly adjusts network quality based on service scenarios to provide high-quality and differentiated computing services.

  • IFIT is deployed in an E2E manner for in-band service flow measurement and fast fault locating.
  • The network uses APN IDs to identify and direct traffic to application-level slices or tunnels, delivering differentiated SLA assurance.

5. Computing security assurance

A security system built upon the network, computing, and security resource pool locates and blocks threats in real time to ensure the trustworthy transmission of computing power.

  • The security system leverages devices, edges, and clouds to ensure secure access to computing resources.
  • The security system provides security resources during computing power scheduling to safeguard computing power transmission.

6. Flexible adjustment of computing power

The network agilely adjusts the supply of computing resources based on real-time changes in the user data volume, providing elastic links between users and computing resources. Computing resources are automatically adjusted based on changes in the user data volume. When the traffic in a region decreases, the computing resources for the region are reduced accordingly, and surplus resources are scheduled to regions with heavy traffic. When the traffic returns to a normal level, the computing resources also automatically return to a normal level.

Differences from Cloud-Network Integration

Cloud-network integration — a concept proposed before computing-network integration — is designed to empower intelligent network innovation and development. Then what is the relationship between the two?

First, computing-network integration is an advanced version of cloud-network integration. Compared with cloud-network integration, computing-network integration promises greater innovation and development space and raises new requirements in terms of evolution objectives, computing power types, and key technologies.

  • In terms of evolution objectives, cloud-network integration focuses on the connection requirements of one-hop cloud access, whereas computing-network integration focuses on closer collaboration between applications and services, aiming to make computing power as conveniently accessible as water and electricity.
  • In terms of computing power types, computing-network integration expands the focus on cloud-based computing power and emphasizes the heterogeneous unification of diversified computing power.
  • In terms of key technologies, the core technology of cloud-network integration is "cloud-based network scheduling", which focuses on resource scheduling. In comparison, computing-network integration emphasizes unified measurement, global scheduling, and elastic orchestration of computing power, extending technological research to application and service levels.
About This Topic
  • Author: Li Yuzhe
  • Updated on: 2023-01-06
  • Views: 3093
  • Average rating:
Share link to