Create a gist now

Instantly share code, notes, and snippets.

What would you like to do?
Note that iostat can overstate bytes written for RocksDB by 2X. I did not adjust the wkb/i, wkb/t and wMB/s numbers
below to account for that. See http://smalldatum.blogspot.com/2016/04/trim-iostat-and-linux.html?fref=gc&dti=489784201120099
- load
ips r/i rkb/i wkb/i Mcpu/i size rss rMB/s wMB/s cpu engine
- old config
57362 0 0 1.14 76 7.8 0.4 0 65.6 4.4 myrocks
73879 0 0 0.74 57 12 27.2 0 54.7 4.2 inno5635
64179 0 0 0.75 63 12 27.8 0 48.2 4.0 inno5717
47901 0 0 0.76 78 12 26.3 0 36.3 3.7 inno802
25747 0 0 1.13 150 11 19.7 0.1 29.1 3.9 toku5717
- new config
54283 0 0.03 1.60 83 14 0.3 0.1 86.7 4.5 myrocks
64402 0 0.02 1.02 72 16 18.4 1.6 65.6 4.6 inno5635
56414 0 0.02 1.03 77 16 17.1 1.4 58.2 4.3 inno5717
42954 0 0 0.02 97 16 16.2 1.0 45.1 4.2 inno802
21611 0 0.02 1.42 179 14 24.6 0.5 30.7 3.9 toku5717
- query @ 12h
tps r/t rkb/t wkb/t Mcpu/t size rss un gn ul gl r/s rMB/s wMB/s cpu engine
- old config
4366 0 0 0.93 811 12 52.8 0.3 0.1 0.5 0.5 0 0 4.0 3.5 myrocks
5651 0 0 7.92 682 19 34.4 0.2 0.1 0.4 0.3 0 0 44.7 3.9 inno5635
5054 0 0 7.89 753 19 32.3 0.2 0.1 0.5 0.3 0 0 39.9 4.0 inno5717
4548 0 0 8.03 796 18 31.9 0.3 0.1 0.6 0.3 0 0 36.5 3.6 inno802
2462 0 0.01 5.07 1398 15 26.1 0.6 0.3 1 0.7 1 0 12.5 3.4 toku5717
- new
tps r/t rkb/t wkb/t Mcpu/t size rss un gn ul gl r/s rMB/s wMB/s cpu engine
5489 0.01 0.07 0.78 642 15 54.1 0.3 0.1 0.5 0.5 25 0.4 4.3 3.5 myrocks
7239 0 0.04 5.15 524 26 29.7 0.3 0.1 0.4 0.2 19 0.3 37.3 3.8 inno5635
6463 0 0.04 5.17 580 26 26.0 0.3 0.1 0.5 0.3 18 0.3 33.4 3.8 inno5717
5855 0 0.05 5.25 623 25 25.7 0.3 0.1 0.6 0.3 17 0.3 30.7 3.6 inno802
3333 0 0 0.08 1043 18 31.5 0.6 0.2 1 0.8 5 0.3 10.0 3.5 toku5717
legend:
* ips, tps - inserts/second, transactions/second
* r/i, r/t - iostat reads per insert and per transaction
* rkb/i, rkb/t - iostat KB read per insert and per transaction
* wkb/i, wkb/t - iostat KB written per insert and per transaction
* Mcpu/i, Mcpu/t - normalized CPU time per insert and per transaction
* r/s, rMB/s, wMB/s - iostat r/s, read MB/s and write MB/s, averages
* size - database size in GB at test end
* rss - mysqld RSS in GB at test end
* cpu - average value of vmstat us + sy columns
* un, gn, ul, gl - 99th percentile response time in milliseconds for UpdateNode, GetNode, UpdateList and GetLinkedList transactions
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment