Dear Mostafa, First, “show vlib graph” describes the entire graph in detail.
Vpp uses ingress flow-hashing (e.g. hardware RSS hashing) across a set of threads running identical graph replicas to achieve multi-core scaling. Historical experiments with pipelining in vpp dissuaded me from pursuing that processing model: the entire pipeline runs at the speed of the slowest stage. More to the point: if the offered workload changes, one needs to reconfigure the pipeline to achieve decent performance. In vpp, you can spin up arbitrary threads and process packets however you like, of course. It would help if you’d describe your application in detail, otherwise we won’t be able to make detailed suggestions. Thanks… Dave From: vpp-dev-boun...@lists.fd.io [mailto:vpp-dev-boun...@lists.fd.io] On Behalf Of Mostafa Salari Sent: Tuesday, October 31, 2017 8:06 AM To: vpp-dev@lists.fd.io Subject: [vpp-dev] VPP default graph Hi, I have 3 issues: 1. I want to know what is the default structure of graph nodes when VPP is running! 2. In dpdk ip_pipeline application, i was able to determine how many instances be created and determine lcore that each instance must run on it! In this way, i was able to make custom optimizations and make a fast packet processing pipeline for my special goal. What is the way in VPP? 3. In order to change the default arrangement, what should i do? Best regards, Mostafa
_______________________________________________ vpp-dev mailing list vpp-dev@lists.fd.io https://lists.fd.io/mailman/listinfo/vpp-dev
