Il 03/07/2012 16:28, Dor Laor ha scritto: >>>> Users using a spinning disk still get IO scheduling in the host though. >>>> What benefit is there in doing it in the guest as well? >>> >>> The io scheduler waits for requests to merge and thus batch IOs >>> together. It's not important w.r.t spinning disks since the host can >>> do it but it causes much less vmexits which is the key issue for VMs. >> >> Does it make sense to use the guest's I/O scheduler at all? > > That's the reason we have a noop io scheduler. But is performance really better with noop? We had to revert usage of QUEUE_FLAG_NONROT in the guests because it caused performance degradation (commit f8b12e513b953aebf30f8ff7d2de9be7e024dbbe). The bio-based path is really QUEUE_FLAG_NONROT++, so it should really be a special case for people who know what they're doing. (Better would be to improve QEMU, there's definitely room for 20% improvement...). Paolo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
