Hi, I'm Paolo, main developer of the BFQ I/O scheduler [1], and this is to share some information on BFQ.
Almost ridiculous to say here, isolation is fundamental with containers. This also applies to I/O bandwidth. In this respect, as many of you probably know, existing techniques for isolating I/O bandwidths cause severe underutilization with common workloads [2] (utilization is often around only 10%). So in the last years I've worked a lot on BFQ to address this issue, and now BFQ recovers this utilization loss almost completely. As a result, BFQ gives from 5x to 10X throughput boost with containers (as well as with multiple groups, virtual machines and so on) [2]. BFQ has been upstreamed since Linux 4.12, so why don't you consider it as a simple solution to greatly boost performance? I'd be happy to help in every respect. Thanks, Paolo [1] https://algo.ing.unimo.it/people/paolo/BFQ/ [2] https://www.linaro.org/blog/io-bandwidth-management-for-production-quality-services/ -- You received this message because you are subscribed to the Google Groups "CoreOS Dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/coreos-dev/2044F4DF-8DB2-4C7A-9336-B41C65809145%40linaro.org.
