Huawei Revealing PCIe SSD for HPC

Huawei Technologies Co., Ltd introduced its latest high IO storage PCIe SSD product for HPC, the ES3000 V3 series card, at SC15.

The ES3000 V3 was on display at the Huawei booth at SC15, along with other HPC solutions including the FusionServer Liquid Cooling System and the OceanStor 9000 big storage System. The ES3000 V3 will be available in the U.S. market and globally in 1Q16.

Huawei also announced that the Poznan Supercomputing and Networking Center (PSNC) in Poland upgraded its HPC data center with a Huawei petaflops-class supercomputer.

The ES3000 V3 NVMe PCIe SSD series can improve storage IO performance to meet big data and virtualization needs, and is compatible with mainstream operation systems and VM systems. It provides up to 3.2TB capacity, up to 3.2GB/s bandwidth and 800,000 IO/s performance, and is for HPC customers in various industries, including finance, transportation, energy, and internet.

Application scenarios for ES3000 series card include:

HPC
Online transaction processing
Online analytical processing
VDI cloud and virtualization
Data catching and indexing

PSNC Challenge
PSNC is affiliated with the Institute of Bioorganic Chemistry Polish Academy of Sciences, serves as a major research center in Poland, and appears on the Top 500 list of the most powerful computing systems in the world. It sought to reach peak supercomputing performance with increased density, while simultaneously achieving dramatic improvements in energy efficiency and reducing OPEX.

Solution
In just three days, Huawei deployed a petaflops-class supercomputer based on the Huawei FusionServer E9000 converged infrastructure blade server chassis with over 1,000 FusionServer CH121 V3 computer nodes. The HPC solution was fully equipped with optimized direct liquid cooling systems, which allowed PSNC to deploy top end performance configuration in a dense blade form factor.

Customer Benefits
PSNC’s new Huawei HPC system is capable of achieving theoretical peak performance of 1.37 petaflops, and averages efficiency of node at 0.901 teraflops. Additionally, the direct cooling system removes up to 80% of heat dissipation, thereby reducing cooling system energy costs by around 40% as compared to traditional air cooling designs. The energy efficiency should amount to millions of dollars in savings by enabling PSNC to deploy higher compute capacity within the same power, cooling and cost parameters.