[vdo-devel] Dedup performance / Writes
Jean-Louis Dupond
jean-louis at dupond.be
Wed Jan 27 22:03:32 UTC 2021
On 27/01/2021 17:38, Gionatan Danti wrote:
> Il 2021-01-27 13:08 Jean-Louis Dupond ha scritto:
>> The write speeds drop to +- 20-50MB/sec. Which is quite low. As native
>> the system can achieve +1000MB/sec (RAID5 on 12 disks).
>> I created the volume with the following arguments:
>> --blockMapCacheSize=1G --sparseIndex=enabled --vdoSlabSize=32G
>
> Do you use an hardware RAID controller? If so, does it have a
> powerloss-protected writeback cache? During the slowdown, what do
> "top" and "iostat -x -k 1" report?
>
HW Raid (dell PERC) with BBU indeed :)
iostat:
avg-cpu: %user %nice %system %iowait %steal %idle
0.76 0.08 6.31 27.42 0.00 65.43
Device r/s w/s rkB/s wkB/s rrqm/s wrqm/s
%rrqm %wrqm r_await w_await aqu-sz rareq-sz wareq-sz svctm %util
sda 6470.00 8742.00 31964.00 58756.00 1092.00 5947.00
14.44 40.49 45.17 0.52 296.76 4.94 6.72 0.07 99.90
dm-0 25.00 0.00 388.00 0.00 0.00 0.00 0.00
0.00 36.36 0.00 0.91 15.52 0.00 36.60 91.50
dm-1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
dm-2 7505.00 14686.00 31504.00 58744.00 0.00 0.00
0.00 0.00 45.87 2.11 375.33 4.20 4.00 0.04 99.80
dm-3 0.00 8575.00 0.00 34816.00 0.00 0.00 0.00
0.00 0.00 8.25 70.78 0.00 4.06 0.08 65.60
dm-4 0.00 0.00 0.00 0.00 0.00 0.00 0.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
sda is the real disk
dm0 is root volume
dm2 is the LV where VDO is on
dm3 is the VDO volume on dm2
top isn't special:
top - 23:02:12 up 2 days, 9:24, 2 users, load average: 5.33, 5.46, 5.32
Tasks: 264 total, 3 running, 261 sleeping, 0 stopped, 0 zombie
%Cpu0 : 1.7 us, 5.1 sy, 0.0 ni, 83.1 id, 8.8 wa, 1.4 hi, 0.0 si,
0.0 st
%Cpu1 : 3.7 us, 5.4 sy, 0.0 ni, 61.8 id, 28.0 wa, 0.7 hi, 0.3 si,
0.0 st
%Cpu2 : 4.4 us, 6.4 sy, 0.0 ni, 73.7 id, 13.8 wa, 1.3 hi, 0.3 si,
0.0 st
%Cpu3 : 2.7 us, 5.7 sy, 0.0 ni, 76.4 id, 14.5 wa, 0.7 hi, 0.0 si,
0.0 st
%Cpu4 : 1.7 us, 5.4 sy, 0.0 ni, 83.8 id, 7.8 wa, 1.0 hi, 0.3 si,
0.0 st
%Cpu5 : 1.4 us, 5.1 sy, 0.0 ni, 64.4 id, 28.1 wa, 1.0 hi, 0.0 si,
0.0 st
%Cpu6 : 2.4 us, 6.7 sy, 0.0 ni, 71.4 id, 14.8 wa, 1.0 hi, 3.7 si,
0.0 st
%Cpu7 : 0.3 us, 5.4 sy, 0.0 ni, 74.6 id, 18.6 wa, 1.0 hi, 0.0 si,
0.0 st
%Cpu8 : 1.0 us, 5.4 sy, 0.0 ni, 59.3 id, 33.2 wa, 1.0 hi, 0.0 si,
0.0 st
%Cpu9 : 0.7 us, 5.1 sy, 0.0 ni, 37.8 id, 55.8 wa, 0.7 hi, 0.0 si,
0.0 st
%Cpu10 : 1.0 us, 5.7 sy, 0.0 ni, 29.3 id, 63.3 wa, 0.7 hi, 0.0 si,
0.0 st
%Cpu11 : 0.3 us, 4.7 sy, 0.0 ni, 73.6 id, 20.3 wa, 1.0 hi, 0.0 si,
0.0 st
MiB Mem : 128568.6 total, 826.3 free, 5691.0 used, 122051.3 buff/cache
MiB Swap: 16384.0 total, 16383.7 free, 0.3 used. 121633.6 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
17353 root 20 0 161056 14052 5044 S 22.6 0.0 157:28.39
sshd
17363 root 20 0 27964 1972 1144 D 12.3 0.0 87:37.24
rsync
1001 root 20 0 0 0 0 S 10.0 0.0 340:33.69
kvdo0:physQ0
998 root 20 0 0 0 0 S 7.6 0.0 258:20.81
kvdo0:journalQ
1000 root 20 0 0 0 0 R 7.3 0.0 259:40.94
kvdo0:logQ0
1002 root 20 0 0 0 0 R 7.3 0.0 241:51.23
kvdo0:hashQ0
1008 root 20 0 0 0 0 S 5.3 0.0 171:23.87
kvdo0:cpuQ0
1009 root 20 0 0 0 0 S 5.3 0.0 171:47.08
kvdo0:cpuQ1
16382 root 20 0 0 0 0 D 3.3 0.0 24:02.53
kworker/u24:2+flush-253:3
1004 root 20 0 0 0 0 S 2.0 0.0 55:23.50
kvdo0:bioQ1
1003 root 20 0 0 0 0 S 1.7 0.0 55:31.11
kvdo0:bioQ0
1006 root 20 0 0 0 0 S 1.7 0.0 55:48.69
kvdo0:bioQ3
1007 root 20 0 0 0 0 S 1.7 0.0 50:05.29
kvdo0:ackQ
1005 root 20 0 0 0 0 S 1.3 0.0 55:22.58
kvdo0:bioQ2
I do a copy from another server, so thats why the rsync and sshd cpu.
More information about the vdo-devel
mailing list