I realize this question isn't specifically fio related, but I imagine many people on this list can accurately answer this question, and I think the answer would be valuable for anyone doing disk IO benchmarking/analysis. I looked for vmstat and iostat mailing lists and documentation and came up short. I apologize if this is just entirely inappropriate for this list. I have noticed that while doing continuous writes with dd (simply streaming from a partition to a file), vmstat's bo (blocks out) counter sits at zero for long lengths of time while iostat's w/s (and related) show significant write traffic. vmstat's bo jumps from 0 occasionally, and I have definitely correlated some of these bo counts to hitting dirty_background_ratio. I have noticed the same behavior when using fio in a write mode with direct=0. In contrast, with fio's direct=1, vmstat's bo always shows non-zero while iostat's w/sec and friends also show non-zero. So, it seems like iostat is reporting all writes to disks/partitions even if they are serviced by cached, but vmstat only reports bo/bi if data is read or written from a disk (ex: when dirty pages are flushed). Is this accurate? Thanks, and again, apologies if this is considered totally inappropriate for this list. Andy -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
