Highlights of the 2022 China Cloud Computing Infrastructure Summit – Launch of Alibaba Cloud Pangu 2.0 and Expert Insights on Smart Computing Infrastructure

The summit held on November 4, 2022 concluded with the official launch of Alibaba Cloud Pangu 2.0, described as a new generation of cloud‑defined infrastructure.

Academician Tao Wenquan delivered a talk on data‑center airflow simulation, covering primary and secondary cooling processes, cabinet numerical models, multi‑scale temperature prediction, rapid airflow organization, and the development of autonomous simulation software.

Alibaba Group Vice President Zhou Ming highlighted the massive scale of Alibaba Cloud’s infrastructure—millions of servers, dozens of regions, hundreds of availability zones—and introduced the seven core systems of Pangu 2.0: Pangu servers, predictable network, data center, intelligent computing Lingjun, liquid cooling, intelligent management, and intelligent operation.

Professor Guo Minyi emphasized the growing gap between AI algorithmic compute demands and chip capabilities, noting trends of irregularity, high concurrency, and multi‑domain fusion, and advocated hardware‑software co‑design to meet future AI workloads.

Senior expert Wang Chao presented Lingjun, a cloud service that enables developers to build enterprise‑grade AI compute centers in two hours, supporting multiple chip architectures and promoting democratized intelligent compute.

Liu Liyin discussed future server trends, predicting servers with 100+ cores that may match the performance of a single CPU, and highlighted challenges in stability, memory, storage, and resource scheduling, proposing a CIPU‑centric, multi‑host, modular, deeply pooled architecture.

Tao Kai introduced storage innovations, including the first‑generation QLC SSD with reduced tail latency and a self‑developed ZNS SSD that integrates mapping table management and garbage collection with distributed storage engines, announcing upcoming cloud‑disk services based on ZNS SSD.

Network architect Xi Yongqing explained that meeting AI compute demands hinges on high‑performance data‑center networking, describing Alibaba’s PredFabric system that offers predictable, low‑latency, cost‑effective interconnect akin to a computer bus.

IDC architect Chao Huaipo outlined four essential capabilities of a smart data center—intelligent interconnect, brain‑driven decision making, executable operations, and observable interaction—enabled by an integrated monitoring architecture.

Operations leader Zhao Xiaoxue addressed the power, cooling, and load challenges of dense compute, emphasizing hardware‑software optimization, carbon‑neutral policies, and a hierarchical compute supply across cloud, edge, and devices.

Technical director Cao Zheng highlighted the evolution from specialized to general‑purpose computing, noting the convergence of CPU and GPU functions as a key driver for future cloud infrastructure upgrades.

The forum concluded with a round‑table discussion titled “Intelligent Computing Frontiers: Infrastructure for the Smart Era,” where experts projected future trends in intelligent computing.