Re: read and write speed.
2011/5/31 djlee064 djlee...@gmail.com: From my (bad) memory, I've tried fsync, and the resulting was still the same in the end, i.e., write 100gb+ and all will be the same. Can't test it now as it is already running other tasks.. I think the small bs (4-8k) results to ~ KB/s level using fsync. and large, 4m+ will get much higher, but at this large bs, eventually, results to the same as no-fsync when you continuously write stretch to 100gb+ Is this too obvious for anyone? (other than the shortstrkoing effect which should be at most 20% diff) Caching effect decreases as the writing size increases. Sure, when you write 100GB+ data, the difference may be subtle. (It still depends on your RAM size.) how then journal-size (e.g., 1gb, 2gb, etc) set by the ceph actually effect in performance (other than reliability and latency, etc) If no reliability is ever needed and, say journal is turned off, how's the performance effect? from my bad memory again, no-journal worsens performance or had no effect . But I think this is again, tested 'not long-enough' e.g., other than journal flushing to disk (which shouldn't be a major bottleneck, as Collin said stuffs in journal gets continuously flushing out to disk. so in other words, set journal to 10gb, and dd for just 10gb, i should get an unbelievable performance, then just change dd to 100gb, how much drop? I haven't try this, but from my very long test wrtites/read, testing many ranges up to 5TB actual content, the disk does about 12.5MB/s at most. For small sets, yes there I see high performance, etc. I am not sure what you mean here. Perhaps somebody can answer it. Sorry. --Henry -- To unsubscribe from this list: send the line unsubscribe ceph-devel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: read and write speed.
On 05/31/2011 01:50 AM, Fyodor Ustinov wrote: Hi! Fresh 0.28.2 cluster. Why reading two times slower than the writing by dd, but rados show different. (Second question - why rados bench crash on read test?) root@gate0:/mnt# dd if=/dev/zero of=aaa bs=1024000 count=1 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 64.5598 s, 159 MB/s root@gate0:/mnt# dd if=aaa of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 122.513 s, 83.6 MB/s root@gate0:/mnt# Additional. fuse vs kernel mount fuse: root@gate0:/mnt# dd if=/dev/zero of=bbb bs=1024000 count=1 conv=fdatasync 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 75.7654 s, 135 MB/s root@gate0:/mnt# dd if=bbb of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 101.613 s, 101 MB/s kernel: root@gate0:/mnt# rm bbb root@gate0:/mnt# dd if=/dev/zero of=bbb bs=1024000 count=1 conv=fdatasync 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 73.143 s, 140 MB/s root@gate0:/mnt# dd if=bbb of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 150.84 s, 67.9 MB/s root@gate0:/mnt# What? Ok. reboot. After reboot: root@gate0:/mnt# rm bbb root@gate0:/mnt# dd if=/dev/zero of=bbb bs=1024000 count=1 conv=fdatasync 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 68.9618 s, 148 MB/s root@gate0:/mnt# dd if=bbb of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 165.564 s, 61.8 MB/s root@gate0:/mnt# Hmm... unmount kernel and mount fuse... root@gate0:/mnt# cd root@gate0:~# umount /mnt root@gate0:~# cfuse -m 10.5.51.230:/ /mnt ** WARNING: Ceph is still under development. Any feedback can be directed ** ** at ceph-devel@vger.kernel.org or http://ceph.newdream.net/. ** cfuse[1102]: starting ceph client cfuse[1102]: starting fuse root@gate0:~# cd /mnt/ root@gate0:/mnt# dd if=bbb of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 93.076 s, 110 MB/s root@gate0:/mnt# What the hell? WBR, Fyodor. -- To unsubscribe from this list: send the line unsubscribe ceph-devel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
read and write speed.
Hi! Fresh 0.28.2 cluster. Why reading two times slower than the writing by dd, but rados show different. (Second question - why rados bench crash on read test?) root@gate0:/mnt# dd if=/dev/zero of=aaa bs=1024000 count=1 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 64.5598 s, 159 MB/s root@gate0:/mnt# dd if=aaa of=/dev/null bs=1024000 1+0 records in 1+0 records out 1024000 bytes (10 GB) copied, 122.513 s, 83.6 MB/s root@gate0:/mnt# root@gate0:/etc/ceph# rados -p test bench 20 write Maintaining 16 concurrent writes of 4194304 bytes for at least 20 seconds. sec Cur ops started finished avg MB/s cur MB/s last lat avg lat 0 0 0 0 0 0 - 0 1 164731 123.966 124 0.445371 0.360663 2 168266 131.967 140 0.160564 0.32065 3 16 11195 126.637 116 0.116967 0.349943 4 16 142 126125.97 124 0.176179 0.369969 5 15 169 154123.17 112 0.12449 0.411138 6 16 202 186 123.971 128 0.175033 0.442003 7 16 241 225 128.541 156 0.163481 0.421575 8 16 271 255127.47 120 0.162152 0.444525 9 16 305 289 128.415 136 0.100893 0.456108 10 16 337 321 128.371 128 0.107163 0.467081 11 16 370 354 128.698 132 0.147602 0.455438 12 16 400 384 127.971 120 0.163287 0.454927 13 16 433 417 128.279 132 0.176282 0.451909 14 16 459 443 126.544 104 3.02971 0.465092 15 16 492 476 126.906 132 0.183307 0.473582 16 16 523 507 126.722 124 0.170459 0.465038 17 16 544 528 124.20884 0.160463 0.462053 18 16 574 558 123.973 120 0.10411 0.478344 19 16 607 591 124.395 132 0.126514 0.48624 min lat: 0.095185 max lat: 3.9695 avg lat: 0.488688 sec Cur ops started finished avg MB/s cur MB/s last lat avg lat 20 16 638 622 124.372 124 2.85047 0.488688 Total time run:20.547165 Total writes made: 639 Write size:4194304 Bandwidth (MB/sec):124.397 Average Latency: 0.513493 Max latency: 3.9695 Min latency: 0.095185 root@gate0:/etc/ceph# root@gate0:/etc/ceph# rados -p test bench 20 seq sec Cur ops started finished avg MB/s cur MB/s last lat avg lat 0 0 0 0 0 0 - 0 1 165842 167.966 168 0.085676 0.279929 2 16 10185169.97 172 0.785072 0.323728 3 16 145 129 171.969 176 0.141833 0.331852 4 16 193 177 176.969 192 0.75847 0.335484 5 15 240 225179.97 192 0.114137 0.332022 6 16 288 272 181.303 188 0.54563 0.339292 7 16 335 319 182.256 188 0.531714 0.341969 8 16 380 364181.97 180 0.101676 0.339337 9 16 427 411 182.634 188 0.216583 0.339264 10 16 471 455 181.968 176 0.803917 0.341281 11 16 515 499 181.422 176 0.112194 0.343552 12 16 559 543 180.968 176 0.241835 0.345668 13 16 600 584179.66 164 0.03 0.347034 read got -2 error during benchmark: -5 error 5: Input/output error root@gate0:/etc/ceph# WBR, Fyodor. -- To unsubscribe from this list: send the line unsubscribe ceph-devel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: read and write speed.
2011/5/31 djlee064 djlee...@gmail.com: A bit off, but can you Fyodor, and all of devs run dd if=/dev/null of=/cephmount/file bs=1M count=1 (10gb) dd if=/dev/null of=/cephmount/file bs=1M count=5 (50gb) dd if=/dev/null of=/cephmount/file bs=1M count=10 (100gb) and continue to 200gb,... 500gb see the MB/s difference, expecting an enormous difference, e.g., starts ~200MB/s and drops down to well less than 50MB/s or even less (Depending on the #disk, but about 12MB/s per-disk is what I have analyzed) I feel that Fyodor's and the rest of you are testing only a very small part, the high rate at the start is likely to be due to the journal size. It's client-side caching effect. Without adding the option conv=fdatasync (or fsync), dd returns the throughput without flushing the data. --Henry -- To unsubscribe from this list: send the line unsubscribe ceph-devel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html