merkushin
commented
10 months ago Pure comments (comment-based repository, not aggregate):
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 44.516 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 73532000 bytes
HTML transferred: 72993000 bytes
Requests per second: 22.46 [#/sec] (mean)
Time per request: 445.159 [ms] (mean)
Time per request: 44.516 [ms] (mean, across all concurrent requests)
Transfer rate: 1613.10 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 10 2.3 10 23
Processing: 227 430 39.9 426 886
Waiting: 225 429 39.9 424 884
Total: 242 440 40.3 436 909
Percentage of the requests served within a certain time (ms)
50% 436
66% 439
75% 441
80% 443
90% 447
95% 458
98% 579
99% 668
100% 909 (longest request)Comment-based aggregate:
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 44.840 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 73532000 bytes
HTML transferred: 72993000 bytes
Requests per second: 22.30 [#/sec] (mean)
Time per request: 448.395 [ms] (mean)
Time per request: 44.840 [ms] (mean, across all concurrent requests)
Transfer rate: 1601.46 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 9 2.3 9 18
Processing: 216 435 91.6 424 1494
Waiting: 215 434 91.7 422 1493
Total: 231 445 91.6 433 1504
Percentage of the requests served within a certain time (ms)
50% 433
66% 436
75% 438
80% 440
90% 444
95% 448
98% 647
99% 879
100% 1504 (longest request)Table-based repository (aggregate):
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 44.132 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 73532000 bytes
HTML transferred: 72993000 bytes
Requests per second: 22.66 [#/sec] (mean)
Time per request: 441.320 [ms] (mean)
Time per request: 44.132 [ms] (mean, across all concurrent requests)
Transfer rate: 1627.13 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 9 2.4 10 22
Processing: 216 427 117.5 411 1533
Waiting: 214 425 117.5 409 1531
Total: 230 436 117.8 420 1545
Percentage of the requests served within a certain time (ms)
50% 420
66% 423
75% 426
80% 427
90% 432
95% 461
98% 639
99% 1532
100% 1545 (longest request)What we can see here:
- Aggregate repositories show worse performance that the "pure" one. I was a bit surprised with it. I didn't expect to see difference between "pure" comments and aggregate comment-based.
- Table-based version had a dramatic decrease of performance in 2% of requests.
- Table-based version still has a little bit better numbers in terms of time for tests (44.132 seconds for table-based vs. 44.516 seconds for "pure" comments) and requests per second (22.66 requests/second for table-based version vs. 22.46 requests/second for "pure" comments).
- Comments-based version use WP caching. Table-based version doesn't use cache yet. When I checked Query Monitor, there were no duplicated queries in comment-based versions and 67 duplicated queries in case of the table-based version. So, I expect we can achieve a dramatic increase of performance for the table-based version by adding caching.
Also have to mention that I posted not the best result for table-based version as I considered it as an extreme and not reliable. However, I got it several times, so I want to share it here:
Concurrency Level: 10
Time taken for tests: 41.934 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 73532000 bytes
HTML transferred: 72993000 bytes
Requests per second: 23.85 [#/sec] (mean)
Time per request: 419.341 [ms] (mean)
Time per request: 41.934 [ms] (mean, across all concurrent requests)
Transfer rate: 1712.42 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 9 2.4 9 22
Processing: 217 406 33.0 402 819
Waiting: 215 405 33.0 401 818
Total: 233 415 33.7 411 840
Percentage of the requests served within a certain time (ms)
50% 411
66% 416
75% 419
80% 421
90% 425
95% 428
98% 452
99% 622
100% 840 (longest request)If we consider longer runs, the difference is less noticeable. (Ran it only once as they take too much time.) "Pure comments":
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 511.681 seconds
Complete requests: 10000
Failed requests: 0
Total transferred: 735320000 bytes
HTML transferred: 729930000 bytes
Requests per second: 19.54 [#/sec] (mean)
Time per request: 511.681 [ms] (mean)
Time per request: 51.168 [ms] (mean, across all concurrent requests)
Transfer rate: 1403.38 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 12 131.1 11 13114
Processing: 213 499 80.5 491 1599
Waiting: 212 498 80.5 490 1597
Total: 228 511 153.8 501 13610
Percentage of the requests served within a certain time (ms)
50% 501
66% 512
75% 520
80% 525
90% 542
95% 565
98% 624
99% 698
100% 13610 (longest request)Table-based aggregate:
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 456.304 seconds
Complete requests: 10000
Failed requests: 0
Total transferred: 735320000 bytes
HTML transferred: 729930000 bytes
Requests per second: 21.92 [#/sec] (mean)
Time per request: 456.304 [ms] (mean)
Time per request: 45.630 [ms] (mean, across all concurrent requests)
Transfer rate: 1573.70 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 13 3.6 13 41
Processing: 221 443 78.7 420 1532
Waiting: 219 441 78.7 419 1530
Total: 238 456 78.5 435 1543
Percentage of the requests served within a certain time (ms)
50% 435
66% 454
75% 473
80% 482
90% 506
95% 544
98% 598
99% 660
100% 1543 (longest request)Comment-based aggregate:
Server Software: nginx/1.25.0
Server Hostname: sensei.test
Server Port: 443
SSL/TLS Protocol: TLSv1.2,ECDHE-RSA-CHACHA20-POLY1305,2048,256
Server Temp Key: ECDH X25519 253 bits
TLS Server Name: sensei.test
Document Path: /?lesson=x-y-z
Document Length: 72993 bytes
Concurrency Level: 10
Time taken for tests: 492.509 seconds
Complete requests: 10000
Failed requests: 0
Total transferred: 735320000 bytes
HTML transferred: 729930000 bytes
Requests per second: 20.30 [#/sec] (mean)
Time per request: 492.509 [ms] (mean)
Time per request: 49.251 [ms] (mean, across all concurrent requests)
Transfer rate: 1458.02 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 3 11 3.9 11 29
Processing: 212 481 80.4 479 1567
Waiting: 211 479 80.4 478 1564
Total: 226 492 80.4 490 1578
Percentage of the requests served within a certain time (ms)
50% 490
66% 497
75% 502
80% 505
90% 514
95% 527
98% 563
99% 660
100% 1578 (longest request)
m1r0
Goal
We need to measure the performance impact when reading data from the new tables.
Scope