nickva commented on PR #5625:
URL: https://github.com/apache/couchdb/pull/5625#issuecomment-3216348604
One missing part of the testing was to see how the PR fares with multiple
dbs when it hits the full cache limit, so I ran the benchmark again (with the
latest version) with 1 database (Q=64, 1M docs) and then 100 dbs also with
Q=64, 10k docs per db. Since the working set is larger and we hit the full
cache a few times, it's less effective, but still shows an improvement both
latency and throughput.
#### MAIN
##### 1 DB
```
HTTP
http_req_duration..............: avg=1.1s min=3.8ms med=1.12s
max=3.97s p(90)=2.32s p(95)=2.52s
{ expected_response:true }...: avg=1.1s min=3.8ms med=1.12s
max=3.97s p(90)=2.31s p(95)=2.52s
{ name:bulk_docs }...........: avg=194.82ms min=122.29ms med=194.57ms
max=506.31ms p(90)=210.71ms p(95)=216.96ms
{ name:get_doc }.............: avg=39.48ms min=3.8ms med=38.29ms
max=3.25s p(90)=50.8ms p(95)=54.52ms
{ name:post_doc }............: avg=1.37s min=39.94ms med=1.28s
max=3.88s p(90)=2.15s p(95)=2.37s
{ name:put_doc }.............: avg=1.91s min=117.86ms med=1.99s
max=3.97s p(90)=2.57s p(95)=2.73s
http_req_failed................: 0.13% 5492 out of 4172744
http_reqs......................: 4172744 4918.641943/s
EXECUTION
iteration_duration.............: avg=4.32s min=1.42s med=4.35s
max=8.74s p(90)=4.97s p(95)=5.17s
iterations.....................: 1390580 1639.152824/s
vus............................: 3175 min=0 max=10000
vus_max........................: 10000 min=10000 max=10000
NETWORK
data_received..................: 2.7 GB 3.1 MB/s
data_sent......................: 738 MB 870 kB/s
running (14m08.4s), 00000/10000 VUs, 1390580 complete and 0 interrupted
iterations
```
##### 100 DBs
```
HTTP
http_req_duration..............: avg=739.75ms min=3.41ms med=722.39ms
max=4.29s p(90)=1.53s p(95)=1.74s
{ expected_response:true }...: avg=739.01ms min=5.91ms med=722.02ms
max=4.28s p(90)=1.53s p(95)=1.74s
{ name:bulk_docs }...........: avg=176.97ms min=104.51ms med=180.09ms
max=222.71ms p(90)=198ms p(95)=204.05ms
{ name:get_doc }.............: avg=72.49ms min=6.42ms med=72.87ms
max=2.5s p(90)=90.76ms p(95)=97.98ms
{ name:post_doc }............: avg=1.01s min=5.91ms med=895.98ms
max=4.28s p(90)=1.6s p(95)=1.8s
{ name:put_doc }.............: avg=1.13s min=57.11ms med=1.02s
max=4.29s p(90)=1.71s p(95)=1.91s
http_req_failed................: 0.11% 6299 out of 5597273
http_reqs......................: 5597273 6640.279304/s
EXECUTION
iteration_duration.............: avg=3.22s min=1.7s med=3.15s
max=7.57s p(90)=4.04s p(95)=4.33s
iterations.....................: 1865324 2212.911958/s
vus............................: 0 min=0 max=10000
vus_max........................: 10000 min=10000 max=10000
NETWORK
data_received..................: 3.5 GB 4.2 MB/s
data_sent......................: 980 MB 1.2 MB/s
running (14m02.9s), 00000/10000 VUs, 1865324 complete and 0 interrupted
iterations
```
#### BTREE CACHE
##### 1 DB
```
HTTP
http_req_duration..............: avg=586.26ms min=3.41ms med=683.49ms
max=3.62s p(90)=1.14s p(95)=1.27s
{ expected_response:true }...: avg=585.63ms min=3.41ms med=682.96ms
max=3.04s p(90)=1.14s p(95)=1.27s
{ name:bulk_docs }...........: avg=118.95ms min=88.45ms med=115.99ms
max=430.22ms p(90)=123.29ms p(95)=125.34ms
{ name:get_doc }.............: avg=12.71ms min=3.41ms med=10.05ms
max=2.11s p(90)=19.59ms p(95)=22.72ms
{ name:post_doc }............: avg=832.66ms min=7.93ms med=808.87ms
max=3.04s p(90)=1.19s p(95)=1.32s
{ name:put_doc }.............: avg=913.62ms min=45.74ms med=896.9ms
max=3.62s p(90)=1.25s p(95)=1.36s
http_req_failed................: 0.09% 6333 out of 6530402
http_reqs......................: 6530402 8459.167081/s
EXECUTION
iteration_duration.............: avg=2.76s min=1.14s med=2.73s
max=6.74s p(90)=3.24s p(95)=3.42s
iterations.....................: 2176466 2819.288849/s
vus............................: 262 min=0 max=10000
vus_max........................: 10000 min=10000 max=10000
NETWORK
data_received..................: 4.1 GB 5.3 MB/s
data_sent......................: 1.1 GB 1.4 MB/s
running (12m52.0s), 00000/10000 VUs, 2176466 complete and 0 interrupted
iterations
```
```
{
"full": {
"value": 0,
"type": "counter",
"desc": "number of times bt_engine cache was full"
},
"hits": {
"value": 37521910,
"type": "counter",
"desc": "number of bt_engine cache hits"
},
"misses": {
"value": 1364431,
"type": "counter",
"desc": "number of bt_engine cache misses"
}
}
```
##### 100 DBs
```
HTTP
http_req_duration..............: avg=578.9ms min=2.82ms med=552.61ms
max=3.93s p(90)=1.25s p(95)=1.44s
{ expected_response:true }...: avg=578.27ms min=3.34ms med=552.29ms
max=3.82s p(90)=1.25s p(95)=1.44s
{ name:bulk_docs }...........: avg=118.11ms min=88.52ms med=117.76ms
max=144.29ms p(90)=126.05ms p(95)=129.4ms
{ name:get_doc }.............: avg=39.67ms min=3.34ms med=38.55ms
max=2.34s p(90)=53.12ms p(95)=58.21ms
{ name:post_doc }............: avg=798.7ms min=6.24ms med=668.4ms
max=3.65s p(90)=1.31s p(95)=1.49s
{ name:put_doc }.............: avg=898.61ms min=6.29ms med=779.32ms
max=3.93s p(90)=1.41s p(95)=1.59s
http_req_failed................: 0.10% 6988 out of 6583385
http_reqs......................: 6583385 8474.816169/s
EXECUTION
iteration_duration.............: avg=2.73s min=1.47s med=2.67s
max=7.45s p(90)=3.45s p(95)=3.71s
iterations.....................: 2194028 2824.380462/s
vus............................: 0 min=0 max=10000
vus_max........................: 10000 min=10000 max=10000
NETWORK
data_received..................: 4.1 GB 5.3 MB/s
data_sent......................: 1.1 GB 1.5 MB/s
```
```
{
"full": {
"value": 726465,
"type": "counter",
"desc": "number of times bt_engine cache was full"
},
"hits": {
"value": 27097580,
"type": "counter",
"desc": "number of bt_engine cache hits"
},
"misses": {
"value": 14718983,
"type": "counter",
"desc": "number of bt_engine cache misses"
}
}
```
<img width="930" height="816" alt="multi_db_btree"
src="https://github.com/user-attachments/assets/ff55f72c-edf9-4d39-abd0-d81292cd8121";
/>
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
