Use InfiniBand RDMA split storage & compute

RDMA supports zero-copy networking by enabling the network adapter to transfer data directly to or from application memory, eliminating the need to copy data between application memory and the data buffers in the operating system. Such transfers require no work to be done by CPUs, caches, or context switches, and transfers continue in parallel with other system operations. When an application performs an RDMA Read or Write request, the application data is delivered directly to the network, reducing latency and enabling fast message transfer.

However, this strategy presents several problems related to the fact that the target node is not notified of the completion of the request (1-sided communications).

Much like other high performance computing (HPC) interconnects, RDMA has achieved limited acceptance as of 2013 due to the need to install a different networking infrastructure. However, new standards^[specify] enable Ethernet RDMA implementation at the physical layer using TCP/IP as the transport, thus combining the performance and latency advantages of RDMA with a low-cost, standards-based solution.^{[citation needed]} The RDMA Consortium and the DAT Collaborative^[1] have played key roles in the development of RDMA protocols and APIs for consideration by standards groups such as the Internet Engineering Task Force and the Interconnect Software Consortium.^[2]

Hardware vendors have started working on higher-capacity RDMA-based network adapters, with rates of 40Gbit/s reported.^[3][4] Software vendors such as Red Hat and Oracle Corporation support these APIs in their latest products, and as of 2013 engineers have started developing network adapters that implement RDMA over Ethernet. Both Red Hat Enterprise Linux and Red Hat Enterprise MRG^[5] have support for RDMA. Microsoft supports RDMA in Windows Server 2012 via SMB Direct.

Common RDMA implementations include the Virtual Interface Architecture, RDMA over Converged Ethernet (RoCE),^[6] InfiniBand, and iWARP.