Directly Access Host Memory Independently of the CPU Even at 10, 40 and 100G Speeds
At low speeds, software alone can perform 100% packet capture but at 10, 40 and 100G, packet loss is inevitable. Unlike standard NIC cards, Accolade’s 1-100GE FPGA-based host CPU offload NIC’s receive and processes packets at the specified line rate (e.g. 100 Gbps) without dropping any packets.
This is due to one of the most beneficial host CPU offload techniques implemented in all Accolade products called direct memory access (DMA). Simply put, DMA is a technique that allows a hardware subsystem such as an FPGA adapter to directly access host memory independently of the CPU. The main advantage of DMA is the host CPU is not burdened with memory transfer and hence is available to perform other tasks. Onboard memory buffers that absorb large bursts of traffic and DMA are key capabilities that enable 100% packet capture on each port.
Compare speed / port configurations across Accolade’s full line of FPGA-based host CPU offload products that enable 100% packet capture, DMA, flow classification, deduplication, packet filtering and much more: