Analysis of the Global HPC Market and Huawei’s High‑Performance Computing Solutions

At the end of the year, Intersect360 released an HPC market analysis report, focusing on Huawei Cloud and AI solutions. The review first outlines the market size and investment, then examines Huawei’s performance and solutions, and concludes with personal viewpoints.

According to Intersect360, the global HPC market (servers, storage, software) reached $356 billion in 2016, a 3.5% increase over 2015, and is projected to grow to $439 billion by 2021 with a CAGR of 4.3%. Servers account for the largest share ($106 billion). The commercial HPC market drives economic growth, and government spending represents 26% of the total, with the U.S. government alone covering about half of the global market.

The fastest‑growing segment is the scalable, web‑oriented Hyperscale market. AI, HPDA, GPU acceleration, and deep learning are the biggest new investment areas in HPC technology. Although public‑cloud consumption of HPC is still small, cloud is the fastest‑growing HPC sub‑market.

Geographically, the U.S. and Canada hold 51% of the HPC market, Europe 24%, and the Asia‑Pacific region 22%, with notable growth in China, Singapore, Australia, and South Korea.

Huawei’s HPC footprint includes:

Supercomputing: the 2 Petaflops “Graham” system built for Compute Ontario, surpassing Poland’s 1 Petaflop “Eagle”.

Automotive: collaborations with ANSYS, ESI, and Altair for aerodynamic and virtual crash testing.

Oil & Gas: certifications with Schlumberger, Halliburton, CGG, and deployments for major national oil companies.

Entertainment: high‑performance rendering for the “Despicable Me” series.

Huawei has established a global Centre of Excellence, including an OpenLab data centre in Munich, to support on‑premise and cloud HPC workloads. Intel processors, development tools, and OmniPath networking are used in over 90% of HPC systems, with installations at universities such as Waterloo, Poznan, EPFL, DTU, and Beijing Jiaotong University.

Key hardware offerings include the FusionServer G5500 (up to 8 NVIDIA Tesla P100 GPUs), the E9000 blade platform, the X6000 high‑density server (supporting up to 24 NVMe drives and 480 k IOPS), and the KunLun Fat‑Node platform (up to 32 CPUs and 24 TB memory). Huawei also provides liquid‑cooling solutions at board and rack levels.

In cloud HPC, Huawei announced HPC Cloud 2.0 at ISC 2017, integrating Mellanox InfiniBand and OpenStack to connect with public clouds such as Azure Stack.

Personal view: Huawei’s HPC portfolio emphasizes compute, with relatively limited focus on storage. While OceanStor 9000 performs well domestically, its overseas presence is modest. Challenges include low adoption of distributed NAS standards, performance limitations of general‑purpose distributed storage, and high space/energy consumption. The shift away from Lustre and GPFS toward BeeGFS, along with the rise of burst‑buffer and all‑flash NAS solutions, presents opportunities for Huawei to develop high‑bandwidth, low‑latency storage offerings.