2

I realise this is probably very basic stuff, but I'm interested to know what you guys would recommend as the best way to recommend server performance? I have just bought a managed dedicated server but I also still have a semi dedicated server running some sites too, so I'm intrigued to see how much improvement there is. Although I expect the dedicated server performance to be better (as the semi-ded is only virtualised), how best should i go about this comparison?

3 Answers 3

3

I use Siege, it's fairly simple to use but fairly powerful. To test it out just do yum install siege (I think it might be in the EPEL repo, alternatively just download the source and compile) then a simple test is

siege -c 64 -t 30s http://domain.com/

where -c 64 is the number of concurrent threads and -t 30s is the time to run the test for. It'll print a load of lines but at the very end you'll get some text like

Lifting the server siege...      done.
Transactions:               1002 hits
Availability:             100.00 %
Elapsed time:              29.76 secs
Data transferred:           3.33 MB
Response time:              1.27 secs
Transaction rate:          33.67 trans/sec
Throughput:             0.11 MB/sec
Concurrency:               42.89
Successful transactions:        1054
/Failed transactions:              0
Longest transaction:           12.89
Shortest transaction:           0.27

which gives you some fairly good comparison info. The key things are Transactions (How many times it managed to make a connection) and Transaction Rate (speed of transactions). Depending on what you're testing for things like Availability (How many of the requests failed) may also be useful. There are quite a few options available so once you've installed just do a siege -h to get all the options.

3
  • out of curiousity, is there any different then apache's AB?
    – grufftech
    May 25, 2011 at 22:51
  • Not a huge difference in that basic example, but you can give siege a text file full of URLs, user agents, post data, cookies etc. so it can simulate real users on your site. I also find siege a bit easier to use and the data output is a bit easier to understand.
    – user80776
    May 25, 2011 at 22:55
  • I should add, one important thing to remember is make sure you're not benchmarking the testing machine. I.E. make sure the machine running siege isn't on 100% CPU, has a good chunk of free RAM and has a low-ish load
    – user80776
    May 25, 2011 at 22:56
1

While it is a little bit older, how about httperf from HP Labs?

http://www.hpl.hp.com/research/linux/httperf/

0

If you are looking for some windows based performance test tools for web servers, I would like to suggest you try WCAT

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .