Red Hat's High Performance Network Add-On delivers remote directory memory access over converged Ethernet (RoCE) for those times when low network latency and high capacity are important. Because RoCE bypasses system and kernel calls to place data directly into remote system memory with less CPU overhead, the High Performance Networking Add-On is ideal for high-speed data processing applications that require low latency, for speeding up cluster locking, or for scaling up applications on distributed systems without investing in specialized networking technologies.
Instead of going through an independent network infrastructure, Red Hat's High Performance Network Add-On for Red Hat Enterprise Linux places data directly into remote system memory using standard 10GigE infrastructure, resulting in less CPU overhead and reduced infrastructure costs.
By using standard RDMA verbs application programming interfaces (APIs), the High Performance Network Add-On is ideal for localized high-speed data processing, such as trading, speeding up I/O, or for scaling up applications on distributed systems without investing in esoteric networking solutions. This means that customers can leverage converged Ethernet network designs to consolidate data and storage networks while reducing their cabling infrastructure, port counts, and associated costs.
RoCE allows customers to maintain a consistent 10-Gigabit Ethernet environment in their datacenters while providing low latency transport for critical transactions. All existing applications using RDMA verbs can leverage a lower cost infrastructure for low latency workloads such as message queues, database transactions, and high-frequency trading applications.