Duo work together to enable NVMe-oF storage solutions

Intel RSD caters to the specific needs of data centre operators, with the ability to efficiently scale and lower operational costs via resource pooling and leverage storage, networking and virtualisation technologies. With resource pooling a key benefit of the Intel RSD architecture, Microsemi provides the fabric interconnect infrastructure with its high density, low power PCIe switches.

The combination of technologies from these three companies provides essential building blocks for original equipment manufacturer (OEM) and original design manufacturer (ODM) customers in the hyperscale data centre and enterprise markets.

“As Intel RSD and AMI’s MegaRAC software are widely adopted by major data centre OEM, ODM and hyperscale customers, we are pleased our innovative Switchtec PCIe switches are an integral part of the RSD ecosystem,” said Pete Hazen, Vice President and Business Unit Manager of storage solutions at Microsemi. “Collaborating with AMI and Intel as part of Microsemi’s Accelerate Ecosystem Programme supports our goal to deliver leading-edge, turnkey solutions for data centre customers worldwide.”

The Intel RSD implementation can utilise NVMe-oF to manage pooled storage for increased performance and efficiency, including higher resource utilisation and lower total cost of operation. Administrators can compose systems from pools of storage tiers to deliver workload-optimised systems for their end users.

For example, an intensive input/output (I/O) workload such as Somatic MUtation FINder (SMuFin) running on multiple servers attached with individual local NVMe storage can now complete the same operation in the same running time on the RSD implementation with a disaggregated NVMe storage model. This can lower storage hardware costs by up to two-thirds. In addition to the key benefits of improved performance and resource utilisation, the combined resources of Microsemi and AMI translate to direct regional support for customers on a global basis.

“Microsemi and AMI continue to be valuable collaborators in bringing OEM solutions based on Intel RSD to market,” said Figen Ulgen, General Manager of Intel Rack Scale Design. “This latest collaboration on NVMe over Fabrics with Intel RSD will offer data centre operators even greater business agility and efficiency.”

OEMs and ODMs can leverage the benefits of combining the AMI MegaRAC development framework with the enhancements from Intel RSD and Microsemi’s Switchtec PCIe switches to lower development costs and speed time to market. Microsemi’s dynamic switch partitioning technology, supported by Switchtec PCIe switches in combination with its Flashtec non-volatile random-access memory (NVRAM) cards and NVMe controllers, enables rapid deployment of NVMe-oF solutions, while significantly offloading central processing unit (CPU) utilisation and practically eliminating system double date rate (DDR) bandwidth requirements. This leads to better overall throughput, latency and quality of service (QoS).

According to G2M Research’s NVMe Ecosystem Market Sizing Report, the NVMe market will be more than $60bn by 2021.

“We are witnessing how NVMe-oF is quickly supplanting traditional methods of storage access. This, combined with the ability to quickly and easily compose and decompose pooled resources via Intel RSD, increases utilisation therefore lowering total cost of operation,” said Subramonian Shankar, President, Founder and CEO of American Megatrends. “As such, we are excited about the capabilities inherent in this combined AMI, Intel and Microsemi solution for the opportunity it represents to better serve new and existing OEM and ODM customers. We also look forward to even greater collaboration between Microsemi and AMI for a variety of product lines including PCIe switches, RAID technology and field programmable gate arrays (FPGAs).”

NVMe-oF solutions are suited for the hyperscale data centre and enterprise markets for NVMe storage pooling applications. The device will also be targeted at applications including extending pooling to other end points, such as FPGAs and general-purpose graphics processing units (GPGPUs).