Arrive’s Storage Acceleration Software

FPGA Acceleration for Storage Applications

The demand for storage disaggregation with NVMe-over-Fabric/NVMe-oF is growing rapidly.  Arrive’s NVMe-oF supports that demand with an NVMe-oF solution using FPGA acceleration.  Arrive’s NVMe-oF solution can provide both popular NVMe-oF techniques: RoCE and NVMe/TCP with high throughput and low latency capabilities.  In addition to the FPGA-based NVMe-oF solution, Arrive’s NVMe-oF solution also provides the software driver/stack that can operate in either full TCP stack offload in an FPGA (TOE offload) or full TCP stack in OS user space (kernel-bypass TCP stack).  With Arrive’s FPGA-based NVMe-oF solution, customers can easily deploy storage disaggregation applications with minimal effort.

 

Features:

  • Up to 40GB/s throughput, 8M IOPS NVMe-oF performance
  • Support two options: NVMe-oF/TCP or RDMA/RCoE
  • With or without TOE (TCP Offload Engine)
  • Compatible with Intel RSD architecture
  • Linux-based software driver for TOE or kernel-bypass TCP stack (without TOE)
  • Plug and play, no customer software change required