Optimizing igb and ixgbe network driver scaling performance

Session information has not yet been published for this event.


One Line Summary

A look at scaling issues found and addressed in the igb and ixgbe drivers, and steps we are exploring to further improve performance.


Multiqueue networking devices are quickly becoming the norm as an ever increasing demand for high bandwidth and low latency have increased the demand on the system beyond what can be handled by a single CPU. In the past additional queues added very little in the way over overall performance gain with many cases showing subpar performance versus a single queue. Some of these issues have been addressed with changes to the network stack itself, however there were multiple issues within the network drivers that were left unaddressed.

I will show the current performance of igb and ixgbe and how their architecture has been updated to improve multiqueue scaling performance. These improvements provide insights into the current status of the multiqueue networking device code, and where future bottlenecks may lie.

Techniques and tools are presented which can be used to gather a picture of the entire network stack and then zoom in on performance bottlenecks. These approaches can be applied to performance analysis beyond just igb and ixgbe and are valuable to audiences both in networking and other communities focused on multiqueue scaling performance.


scaling, kernel

Presentation Materials



  • Biography

    Alexander Duyck is a senior network software engineer at Intel where he works on network device drivers and supporting new network device features within the Linux network stack.