[Edited Message Follows] We are trying to build an application which communicates with vpp over memif interface. VPP has 8 worker threads, and IP address have been pinned to each of these threads (via rx-placement). Out application connects to VPP using memif (VPP Master & App Slave), with 8 queues.
On our application, we have 8 worker threads, reading from one of the 8 memif queues, such that · rx-queue-1 is on thread 1 · rx-queue-2 on thread-2 and so on. Each of these thread will write to the associated queue-id of the memif rx, e.g. · thread-1 writes on tx-queue-1 · thread-2 writes on tx-queue-2 >From another post in this mail list we understand that the tx-queues are >placed in sequence of the worker threads, like tx-0 on main thread, tx-1 on >wk_1, tx-2 or wk-3 and so on. So is it safe to assume that if we place our >memif rx-queues respectively in the same sequence, then the packet belonging >to an IP address will land on same thread, e.g IP_1 à vpp_wk_1 à memif_tx_1 : memif_rx_queue_1 à app_thread 1 à memif_tx_queue_1 : memif_rx_1 à vpp_wk_1 Is this understanding correct ? If I am missing something else, or there is a better way to achieve this, please suggest. Thanks, Swarup.
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#20015): https://lists.fd.io/g/vpp-dev/message/20015 Mute This Topic: https://lists.fd.io/mt/85132356/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
