Hi, I'm currently compare results of "rados bench" vs fio-librbd, and I notice that with rados bench (4k randread, 1osd) "rados bench -p pooltest 300 rand -t 128" Bandwidth (MB/sec): 284 Average IOPS: 72750 Increasing threads number don't improve results after 128 threads but launching the same rados bench twice in parallel, give me rados bench1: Bandwidth (MB/sec): 247 Average IOPS: 63203 rados bench2: Bandwidth (MB/sec): 251 Average IOPS: 64175 Is it expected ? If I compare with fio, it seem like that threads number is related to iodepth in fio with fio, I got around 70000 iops with 1 job iodepth=128 121000 iops with 2 jobs iodepth=128 Could it be possible to improve rados bench to run it multiple time in parallel and aggregate results ? It could be usefull to compare with fio. Regards, Alexandre -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
