100GbE Networking Improvements On Tap For Linux 5.3
While we are just at the RC2 stage for the Linux 5.2 kernel, already queuing in net-next for Linux 5.3 are some 100GbE networking driver improvements.
Intel's ICE 100GbE wired network driver is among the high-speed LAN drivers seeing improvements for the next kernel. The "ICE" driver has VF structure optimizations to use less memory, more efficient ordering for transmit buffer and ring structures, and other enhancements so far. Some of those details here and expect more ICE driver work over the weeks ahead given how young the cycle is until Linux 5.3's merge window in July.
Meanwhile Qlogic's QED 100G driver has also seen improvements and arguably more prominent/ The QED driver is seeing improved performance on 100G links for offload protocols. "This patch series modifies the current implementation of PF selection. The refactoring of the llh code enables setting additional filters (mac / protocol) per PF, and improves performance for offload protocols (RoCE, iWARP, iSCSI, fcoe) on 100G link (was capped at 90G per single PF). Improved performance on 100G link is achieved by configuring engine affinty to each PF."
Great to see all the Linux networking performance improvements happening across many drivers. Also queued in net-next for Linux 5.3 is a lot of new "nexthop" code including a new user-space API around next hops but haven't yet looked much into it to see how it will improve Linux's next-hop routing.
Intel's ICE 100GbE wired network driver is among the high-speed LAN drivers seeing improvements for the next kernel. The "ICE" driver has VF structure optimizations to use less memory, more efficient ordering for transmit buffer and ring structures, and other enhancements so far. Some of those details here and expect more ICE driver work over the weeks ahead given how young the cycle is until Linux 5.3's merge window in July.
Meanwhile Qlogic's QED 100G driver has also seen improvements and arguably more prominent/ The QED driver is seeing improved performance on 100G links for offload protocols. "This patch series modifies the current implementation of PF selection. The refactoring of the llh code enables setting additional filters (mac / protocol) per PF, and improves performance for offload protocols (RoCE, iWARP, iSCSI, fcoe) on 100G link (was capped at 90G per single PF). Improved performance on 100G link is achieved by configuring engine affinty to each PF."
Great to see all the Linux networking performance improvements happening across many drivers. Also queued in net-next for Linux 5.3 is a lot of new "nexthop" code including a new user-space API around next hops but haven't yet looked much into it to see how it will improve Linux's next-hop routing.
Add A Comment