Direct Memory Access (DMA)
Advanced FPGA-based Host CPU Offload
After all packet processing concludes, an ANIC adapter efficiently transfers all relevant packets and associated packet descriptors (metadata) across the PCIe bus directly in to host memory for consumption by the host application.
The main advantage of DMA is the host CPU is not burdened with memory transfer and hence is available to perform other more important tasks.
Multi-core DMA is a technique that makes processing by multiple host CPU cores more efficient. For example, assume the host Intel CPU has 4 cores (up to 64 cores are supported) with each operating independently of the other 3 cores. The ANIC adapter is programmed to write data in to 4 independent host packet buffers (HPB) and each CPU core (and related application thread) is in turn programmed to process only its corresponding HPB. In this way a network security or monitoring application can take advantage of parallel processing of data thus achieving higher levels of speed and efficiency.
PF_RING is an open source, Linux kernel module for packet capture that is supported by most open source IPS/IDS solutions such as Snort, Suricata, and Bro as well as other applications such as Wireshark and Argus. PF_RING has native support for Accolade ANIC adapters which makes integration with any application that uses PF_RING seamless.