Building upon last-year's record-breaking three million I/O per second Phase Change Memory (PCM) demonstration, HGST, in collaboration with Mellanox Technologies, is showcasing a PCM-based, RDMA-enabled in-memory compute cluster architecture that delivers DRAM-like performance at a lower cost of ownership with greater scalability. While modern data center applications can benefit from more main memory, today's DRAM approaches are expensive to scale because of that memory's volatility: DRAM stores data in leaky capacitors, and thus needs to be rewritten many times per second to stave off data loss. This refresh power consumption can be as much as 20-30% of the total server energy4. Emerging non-volatile memory technologies, such as PCM, do not have this refresh power demand thereby enabling far greater scalability of main memory than DRAM.

HGST says its new persistent memory fabric technology delivers reliable, scalable, low-power memory with DRAM-like performance, and does not require BIOS modification nor rewriting of applications. Memory mapping of remote PCM using the Remote Direct Memory Access (RDMA) protocol over networking infrastructures, such as Ethernet or InfiniBand, enables a wide scale deployment of in-memory computing. This network-based approach allows applications to harness the non-volatile PCM across multiple computers to scale out as needed.

The HGST/Mellanox demonstration achieves random access latency of less than two microseconds for 512 B reads, and throughput exceeding 3.5 GB/s for two KB block sizes using RDMA over InfiniBand.