Hacker News new | past | comments | ask | show | jobs | submit login

You also tend to end up dedicating cores. That is a side effect of spinning in a loop checking for the next packet/queue processing rather than waiting for an interrupt.

There is also the additional overhead of VFIO/IOMMU's which can easily eat double digit perf vs doing the processing in a trusted kernel context.

So with a normal machine the cores/etc are balanced between processes, with DPDK/etc thats hard as some of the cores will be 100% consumed in those spinloops even if your experiencing 5% line rate.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: