[Popcon-developers] popcon issues
Petter Reinholdtsen
pere at hungry.com
Mon Jun 22 12:00:29 UTC 2015
[Bill Allombert]
> Not about duplicated UID.
>
> However I start to believe that the issue is that the webserver is too
> slow to answer:
Also, according to siege it can handle around 18 transactions per second
for this static page.
% siege -t20s http://popcon.debian.org/
** SIEGE 2.70
** Preparing 15 concurrent users for battle.
The server is now under siege...
Lifting the server siege... done.
Transactions: 363 hits
Availability: 100.00 %
Elapsed time: 19.94 secs
Data transferred: 1.19 MB
Response time: 0.33 secs
Transaction rate: 18.20 trans/sec
Throughput: 0.06 MB/sec
Concurrency: 6.09
Successful transactions: 363
Failed transactions: 0
Longest transaction: 0.36
Shortest transaction: 0.31
[...]
Starting the submit script is a lot slower:
% siege -t20s http://popcon.debian.org/cgi-bin/popcon.cgi
** SIEGE 2.70
** Preparing 15 concurrent users for battle.
The server is now under siege...
Lifting the server siege... done.
Transactions: 144 hits
Availability: 100.00 %
Elapsed time: 19.32 secs
Data transferred: 0.02 MB
Response time: 1.41 secs
Transaction rate: 7.45 trans/sec
Throughput: 0.00 MB/sec
Concurrency: 10.53
Successful transactions: 144
Failed transactions: 0
Longest transaction: 1.79
Shortest transaction: 0.71
[...]
This mean at best, around 26000 submissions can be received in an hour
or around 450 in a minute. In reality it is slower to process a
submission than to reject an invalid one, so the real capasity will be
lower.
I tested a real submission using 'time', and it used 0.83 second to
deliver my popcon report via http, on average over 3 deliveries. That
is in the range listed above (0.71 - 1.79 s).
Perhaps it is time to cluster the reception of requests? How are Ubuntu
able to handle the load they are receiving? Should we use their script
instead?
--
Happy hacking
Petter Reinholdtsen
More information about the Popcon-developers
mailing list