For siege, you would: From a

Shaggy — Wed, 21 Oct 2009 14:44:13 +0000

For siege, you would:
From a client that is not on the server...
1) generate a urls.txt file, containing the URLs you want visited by the utility

2) Decide on the number of simultanious users you want visiting your site

3) How may pages each user is going to visit...

So, if you wanted 20 users to visit 10 pages you would:

siege -c20 -r10 -f ./urls.txt

The output will be something like:
The server is now under siege...done
Transactions: 200 hits
Availability: 100.00 %
Elapsed time: 12.76 secs
Data transferred: 1.49 MB
Response time: 0.65 secs
Transaction rate: 15.68 trans/sec
Throughput: 0.12 MB/sec
Concurrency: 10.19
Successful transactions: 200
Failed transactions: 0
Longest transaction: 1.91
Shortest transaction: 0.51

While siege is running, watch the output of vmstat on your VPS, as well as top to see what processes are doing while under the load. You may want to specify more requests to siege so the test takes longer - and you watch more data on the server

Some reading is required to interpret vmstat output, but the first three columns are particularly interesting (r, b, w)

The ideal value for each is '0' - as your server shows duress, these numbers will grow as more and more processes stack up waiting for the bottlenecked resource(s)

r = # of processes in the run queue (CPU?!)
b = processes blocked up because they're waiting for IO (disk?!)
w = runnable but swapped - (memory?!)

Be careful about where you are performing the test too - it is possible to saturate your network connection, which will bottleneck your test!

Cheers,
Shaggy.

The test environment is on a

toughntwisted — Wed, 21 Oct 2009 14:16:15 +0000

The test environment is on a VPS server I pay for monthly, How would I use these programmes to test the load capabilitys?

Sam

Hi There; Do you have a

Shaggy — Wed, 21 Oct 2009 13:32:29 +0000

Hi There;

Do you have a development / test environment?
Populate the environment, note the specifications, and then do a little load test using 'seige', 'ab', or similar with some traffic parameters you're expecting to get first off, and then increase from there to see where things fall over.

If the application is decently tuned, depending on how your webserver / PHP are configured, I'd expect you to hit the RAM wall first. Once over that hurdle, disk IO is usually next, especially if your database is sitting on the same machine.

Cheers,
Shaggy.