Difference between revisions of "VPP/FAQ"
From fd.io
< VPP
(Created page with "== Frequently Asked Questions == # Q: I'm using multiple vpp worker threads, yet only one of them is processing traffic. Why? <br /> A: Hardware RSS hashing sends all traffi...") |
(→Frequently Asked Questions) |
||
| (One intermediate revision by one other user not shown) | |||
| Line 1: | Line 1: | ||
| + | __TOC__ | ||
== Frequently Asked Questions == | == Frequently Asked Questions == | ||
| − | + | === Q: I'm using multiple vpp worker threads, yet only one of them is processing traffic. Why? === | |
| + | <br /> A: Hardware RSS hashing sends all traffic from a given flow to the same queue. Each queue is assigned to a single worker thread. Make sure that you're sending traffic from a reasonably large number of flows. | ||
| + | === Q: Why do I see RX misses on an interface at a low data-rate? === | ||
| + | <br /> A: Check to make sure that the Linux kernel isn't scheduling other work on the same core(s) as vpp. It just won't do for vpp to take a forced 10ms vacation in the middle of packet processing. Check out the vpp autoconfiguration tool. | ||
Latest revision as of 15:34, 29 November 2017
Contents
Frequently Asked Questions
Q: I'm using multiple vpp worker threads, yet only one of them is processing traffic. Why?
A: Hardware RSS hashing sends all traffic from a given flow to the same queue. Each queue is assigned to a single worker thread. Make sure that you're sending traffic from a reasonably large number of flows.
Q: Why do I see RX misses on an interface at a low data-rate?
A: Check to make sure that the Linux kernel isn't scheduling other work on the same core(s) as vpp. It just won't do for vpp to take a forced 10ms vacation in the middle of packet processing. Check out the vpp autoconfiguration tool.
