Views: 42 Author: Site Editor Publish Time: 2022-04-01 Origin: Site
With the advent of AI and big data era, the business of data centers is constantly evolving and new technologies are constantly being incorporated into data centers. This includes two implications of an intelligent data center.
On the one hand, it is how to optimize data center operation based on massive data and artificial intelligence technology; on the other hand, data centers will increasingly carry big data business, artificial intelligence training scenarios and artificial intelligence application scenarios, in which data centers themselves need to adapt to the new intelligent business needs.
In the current stage, in addition to the traditional CPU-centric data center, the needs of AI will also be considered, such as: building resource pools for joint CPU and GPU and building resource pools for carrying out computing such as FPGAs for edge computing scenarios. In this process, AI-driven data center energy saving, AI-driven automatic optimization and intelligent operation of computing, storage and network, and AI-driven inspection robots, etc., especially the monitoring and troubleshooting of some faults in carrier-specific NFV scenarios, are the new demands and challenges that the current intelligent data center is facing.
In the next stage, higher demands for unification and convergence will be raised, including the unification of edge and core and the convergence, standardization and IoT of AI and various systems. In particular, the further unification and convergence of the entire data center at the edge, even between the devices at the edge and the data center at the edge, is also something we need to focus on in the future.
And in the final stage, it is hoped that a fully automated data center can be realized.
For operators, there are many challenges in building intelligent data centers at this stage, including the transformation of infrastructure, how to adapt to some new business needs of artificial intelligence and big data, how to provide richer API interfaces and more data storage.
First, the first aspect is intelligent data center energy-saving technology. Many experts have also talked about the possibility of developing many energy-saving technologies at the physical infrastructure level and introducing many energy-saving related devices to reduce the energy consumption of our data centers.
The second case is server customization, which is in line with the evolution of intelligent data centers. Among them, China Telecom's early customization was whole cabinet servers and standalone servers, and in 2015 and 2016 it was hyper-converged customization and low-power customization servers. In 2017, the main ones considered were customized servers in the ServerSAN area, customized servers for NFV, and customized servers for GPU for artificial intelligence.
As the data center business evolves, the server level must adapt to the corresponding changes and carry out new types of server customization work. This work and ODCC's work are complementary and mutually reinforcing.
The third aspect is to build a PaaS platform for artificial intelligence in the data center, which is currently being tried out mainly in China Telecom's cloud computing lab. There are two types of AI-oriented PaaS platforms: one for the public cloud and one for the industry.
The fourth aspect is AI-assisted intelligent operation and maintenance. Currently, the original operation and maintenance method is facing many challenges, such as: the IT architecture after virtualization, end-to-end operation and maintenance tools across computing, storage and network, the application of containers, microservices and virtualization, and multi-vendor integration.
We try to build an AI-assisted operation and maintenance system and study how to make full use of big data and artificial intelligence technologies from the data-aware level, fault diagnosis level, fault prediction and fault self-healing level to make the whole data center operation and maintenance work more intelligent and automated.
Of course, the road of data center intelligence has just begun, and there is still a lot of work to be done in the future, and the industry needs further research and cooperation, so we believe that the future data center can have higher intelligence.