Need Help?

Huawei and Mellanox Preview Industry’s First NVMe-oF™-based Resource Pooling Solution

2016/9/1 0:00:00 Author: Source:

[Shanghai, China, September 1, 2016] Huawei, a leading global information and communications technology (ICT) solutions provider, and Mellanox previewed the industry’s first resource pooling solution based on NVM Express® over Fabrics (NVMe-oF™) at HUAWEI CONNECT 2016. According to test statistics, the solution delivers higher storage bandwidth and performance, and boosts the random read/write performance by at least 2,000 times compared with the traditional SAS solution.

Server booth at HUAWEI CONNECT 2016

As a leading global supplier of end-to-end InfiniBand and Ethernet interconnect solutions and services for servers and storage, Mellanox, in cooperation with Huawei, built the industry’s first resource pooling solution based on NVMe-oF™. The solution aims to attain lower resource access latency and higher performance as well as a more robust framework for NVMe SSDs, thus providing more agile user experience.

“NVMe-oF is designed to set a new standard in high bandwidth, low latency and high IOPS storage,” said Michael Kagan, CTO of Mellanox Technologies. “The solution we have designed with Huawei is designed to enable effortless communication for servers and solid state disks (SSDs) while offering high bandwidth and low latency through Remote Direct Memory Access (RDMA).”

“NVMe has brought new changes and challenges to the computing architecture, taxing the traditional architecture beyond its capacities,” said Qiu Long, President of Huawei IT Server Product Line. “Huawei servers offer outstanding computing platforms and high-performance NVMe SSDs. We partner with Mellanox this time to present the cutting-edge NVMe over Fabrics solution to address customers’ requirements for high-performance storage. In the future, Huawei will work with more partners to forge an NVMe over Fabrics ecosystem, to accelerate the process that makes NVMe over Fabrics the mainstay for data center resource pool construction.”

In the solution, Huawei FusionServer E9000 blade servers are used to build the hardware platform, Huawei CE8860 100G Ethernet switches are used to build the network hardware platform, and Mellanox ConnectX-4 100G Ethernet cards are used as cluster interconnect cards.

Huawei’s FusionServer E9000 converged architecture blade server strives to converge infrastructure by integrating compute, storage and network. This cohesive approach improves operational efficiency and data-handling capacity for customers. E9000 supports full NVMe SSD flash blades, which, in combination with the high-speed, low-latency bandwidth of 100G networks, enable a single-blade throughput rate of 10 million IOPS, unleashing massive potential of services.

Huawei CE8860 is a new-generation Ethernet switch developed by Huawei for data centers. It features high performance, high density, low latency, and flexible plug-in cards. A CE8860 supports up to 32 100GE, 64 40GE, 128 25GE, or 128 10GE ports. Combining flexible plug-in cards in CE8860 can build diversified high-density access and aggregation capabilities.

Mellanox’s ConnectX®-4 VPI provides 10, 25, 40, 50, 56, or 100 Gbit/s throughput capacity, supports standard InfiniBand and Ethernet protocols, and offers excellent flexibility. The ConnectX-4 can handle 150 million messages per second, with a latency of only 0.7 µs, while integrating RDMA, GPUDirect, SR-IOV and other turbo caches. This enables the most efficient computing and storage platforms.