How to get more requests per second? Is there a plateau/ceiling?

jaydee

I'm coming from shared hosting and I'm testing how fast WordOps on a VPS can go. I used Apache Bench for testing, like this:

abs -n 100 -c 10 <SITE_URL>
abs -n 500 -c 25 <SITE_URL>
abs -n 1000 -c 50 <SITE_URL>
abs -n 1500 -c 75 <SITE_URL>
abs -n 1000 -c 200 <SITE_URL>
abs -n 1000 -c 400 <SITE_URL>

I started out with the most basic 1 vCPU 1GB memory DO VPS, with no caching. This got me roughly 11 requests per second on all test. Pretty cool, my shared host got 8-9, so a 29% increase.

Then I enabled fastcgi_cache. Performance jumped up to 78 requests per second. Awesome!

I assumed this was a hardware limit - because it's only a $5 VPS, so I upgraded to 1 vCPU 2 GB memory, and then 2 vCPU 2 GB memory. To my surprise, the performance remained the same! What's going on here? Do I need to reconfigure something? Is this a software limit?

Hope you can help.

nsgoyat

jaydee I don't know if this will help in improving or not but the thing is WordOps optimizes MySQL my.cnf file as per the available resources (mostly RAM).

So, if you upgrade the RAM then try to run the following command:
wo stack upgrade --mysql --force

This way, WordOps will re-optimize the server's MySQL Configuration according to the available RAM size.

Following the thread. I would love to see other people's insights over this issue.

jaydee

nsgoyat Interesting. I should've done another bench before I ran the commands. I read up on wo stack and ran wo stack upgrade which seemed to check/upgrade/update all the elements in the stack. The new bench got me 105 requests per second. Higher, but I expected more from doubling the CPU and RAM. I also did wo stack upgrade --mysql --force but it didn't change.

@jond1 Just to clarify, I'm not looking for higher performance on its own. I'm trying to figure out what elements control the performance. (Your 1k is really sweet tho 😉 )

jond1

@jaydee I ran a test against my production site, which is running on the lowest option for Lightsail ($3.50/mo), here are my results. You'll see, 350 requests per second, and the server never went above 0.5 load.

ab -n 10000 -c 400 https://www.jonathan.vc/
This is ApacheBench, Version 2.3 <$Revision: 1843412 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking www.jonathan.vc (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests


Server Software:        cloudflare
Server Hostname:        www.jonathan.vc
Server Port:            443
SSL/TLS Protocol:       TLSv1.2,ECDHE-ECDSA-AES128-GCM-SHA256,256,128
Server Temp Key:        ECDH P-256 256 bits
TLS Server Name:        www.jonathan.vc

Document Path:          /
Document Length:        65572 bytes

Concurrency Level:      400
Time taken for tests:   28.705 seconds
Complete requests:      10000
Failed requests:        763
   (Connect: 0, Receive: 0, Length: 763, Exceptions: 0)
Total transferred:      665490763 bytes
HTML transferred:       655720763 bytes
Requests per second:    348.37 [#/sec] (mean)
Time per request:       1148.202 [ms] (mean)
Time per request:       2.871 [ms] (mean, across all concurrent requests)
Transfer rate:          22640.39 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:       26   45  40.9     36     301
Processing:   597 1073 275.4    961    2538
Waiting:      283 1035 272.6    921    2516
Total:        628 1118 296.5    998    2566

Percentage of the requests served within a certain time (ms)
  50%    998
  66%   1046
  75%   1110
  80%   1281
  90%   1645
  95%   1768
  98%   2026
  99%   2106
 100%   2566 (longest request)

VirtuBox

Hello jond1,
For me there is something wrong in your comparison, because in your first benchmark, there were 665MB transferred with a transfer rate of 22MB/s.
But in the second one, there only were 3,97MB transferred with a transfer rate of 401KB/s.

jond1

I ran the same test against another production site I have running in a high availability cluster, using EFS, Redis, and RDS, all within AWS. Here are the results, 1,000 requests per second and no server impact to the load.

Concurrency Level:      400
Time taken for tests:   9.664 seconds
Complete requests:      10000
Failed requests:        0
Non-2xx responses:      10000
Total transferred:      3970000 bytes
HTML transferred:       0 bytes
Requests per second:    1034.77 [#/sec] (mean)
Time per request:       386.559 [ms] (mean)
Time per request:       0.966 [ms] (mean, across all concurrent requests)
Transfer rate:          401.18 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:       34  257  48.0    269    1198
Processing:    41  121  40.0    128     317
Waiting:       15   66  28.4     55     303
Total:         80  377  31.6    367    1363

Percentage of the requests served within a certain time (ms)
  50%    367
  66%    371
  75%    375
  80%    381
  90%    427
  95%    447
  98%    473
  99%    482
 100%   1363 (longest request)

jond1

@jaydee I'm running stock install for the first set of results. Second set is customized for high availability.

To note, I'm running php74 on all my sites. The first set of results, I'm running with fastcgi since it's on a single OS (MySQL, filesystem, etc).

Generally speaking, you will get more performance out of more hardware, however I also designed my infrastructure to be highly available and fault tolerant across multiple AWS zones. You should be able to see a lot of performance out of a single instance though. The smallest instance on DO will likely not give the greatest performance.

I suggest trying the 4gb, 2 CPU for $20/mo. That's a steal of a price for that performance.

Also make sure you have SWAP setup. AWS instances don't automatically have those, I'm not sure about DO, it's been a while since I've used their products.

jaydee

jond1 I tried 2 CPU 4GB RAM but I remain at 105 RPS. 😞

My test site is https://one.toscani.dev/. The theme is custom made from scratch. Very small considering the large images. The only external call is to Typekit. The domain is at Namecheap, the DNS is at Cloudflare (free), and as mentioned before the hosting is on DO, and the install is --wpfc with just the Nginx Helper plugin.

jond1

@VirtuBox that was because of a gallery on the homepage, lots of big images.

splaquet

I've been trying to figure out what's up with our install. I ran into this (older) post, but figured I'd give it a go.

We're experiencing something that's essentially clogging our EC2 instance. We currently have about 6 EC2s connected to 1 RDS. All of the other sites are fine, leading me to assume that the problem is isolated to the 1 EC2 instance. And, the huge variable on that instance is that it has WooCommerce installed.

The install is still in it's DEV stages, so it's really not seeing any type of real traffic... but how/why has it been locking up and coming to a stand still?

Also worth noting, we didn't seem to experience this when the build was isolated to a single stack instance. It only started generating issues when we migrated to the RDS. ...and we've optimized as many of the RDS settings as we're able to... so, now we're at a stand still

From what I can tell, this error would imply that I'm running into an OpenSSL issue... but, I'm at a loss as to how to fix. I've done a bunch of research and everything related to OpenSSL has been updated... has anyone else experienced anything like this?

And outside of the ApacheBench box, any other ideas or suggestions on what might be troubling our install?

(this first part repeats A LOT... so, I'm omitting a few thousand lines)

$ ab -n 10000 -c 400 https://domain.tld/
This is ApacheBench, Version 2.3 <$Revision: 1879490 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking humphreysusa.net (be patient)
SSL read failed (1) - closing connection
804301B8EF7F0000:error:0A000197:SSL routines:SSL_shutdown:shutdown while in init:../ssl/ssl_lib.c:2244:
804301B8EF7F0000:error:0A000126:SSL routines:ssl3_read_n:unexpected eof while reading:../ssl/record/rec_layer_s3.c:308:
SSL read failed (1) - closing connection
804301B8EF7F0000:error:0A000197:SSL routines:SSL_shutdown:shutdown while in init:../ssl/ssl_lib.c:2244:
804301B8EF7F0000:error:0A000126:SSL routines:ssl3_read_n:unexpected eof while reading:../ssl/record/rec_layer_s3.c:308:
SSL read failed (1) - closing connection
804301B8EF7F0000:error:0A000197:SSL routines:SSL_shutdown:shutdown while in init:../ssl/ssl_lib.c:2244:
804301B8EF7F0000:error:0A000126:SSL routines:ssl3_read_n:unexpected eof while reading:../ssl/record/rec_layer_s3.c:308:
Completed 10000 requests
Finished 10000 requests

Server Software:        cloudflare
Server Hostname:        domain.tld
Server Port:            443
SSL/TLS Protocol:       TLSv1.2,ECDHE-ECDSA-CHACHA20-POLY1305,256,256
Server Temp Key:        X25519 253 bits
TLS Server Name:        domain.tld

Document Path:          /
Document Length:        0 bytes

Concurrency Level:      400
Time taken for tests:   69.116 seconds
Complete requests:      10000
Failed requests:        10000
   (Connect: 0, Receive: 0, Length: 10000, Exceptions: 0)
Total transferred:      1665265575 bytes
HTML transferred:       1657152949 bytes
Requests per second:    144.69 [#/sec] (mean)
Time per request:       2764.626 [ms] (mean)
Time per request:       6.912 [ms] (mean, across all concurrent requests)
Transfer rate:          23529.20 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        7  156 170.7    113    1021
Processing:   199 2024 539.5   1974    4820
Waiting:       18  627 659.7    379    3313
Total:        217 2180 634.6   2088    5740

Percentage of the requests served within a certain time (ms)
  50%   2088
  66%   2187
  75%   2335
  80%   2430
  90%   2795
  95%   3020
  98%   4575
  99%   5093
 100%   5740 (longest request)