Load testing a website?
March 14, 2014 8:13 PM Subscribe
What's a good free or cheap service for load/stress testing a website?
ProxySniffer to stress and NewRelic to analyse the site.
posted by oxit at 1:21 AM on March 15, 2014 [1 favorite]
posted by oxit at 1:21 AM on March 15, 2014 [1 favorite]
Installing siege is free, really simple and can get you quite a long way.
posted by gregjones at 4:49 AM on March 15, 2014
posted by gregjones at 4:49 AM on March 15, 2014
siege and ab are fine for heavy unsubtle whacks at server infrastructure, but they're not great (ab especially) at simulating real-world load, load from different geographical regions, assessing usage patterns (anon vs logged in, authenticated sessions, SSL load) or checking performance of client-driven functionality like Javascript. For all that, it's worth going with a cloud-based approach, recording sessions with something like BlazeMeter's JMeter scripting proxy extension and cloning them out into user scenarios.
nthing NewRelic for analysis.
posted by holgate at 12:15 PM on March 15, 2014
nthing NewRelic for analysis.
posted by holgate at 12:15 PM on March 15, 2014
This is going to sound weird, but my first thought was that you can think of something socially inflammatory and post it in a very active subReddit and put a link to your site as the "proof" and then see if tons of people click through. See The Reddit Effect. Not sure how effective it will be, but it's cheap.
posted by CathyG at 4:54 PM on March 15, 2014
posted by CathyG at 4:54 PM on March 15, 2014
I've always been partial to JMeter. I'll take a set of real world traffic logs and use that to create test plans that hit those same patterns much much more frequently.
Of course, if there are specific things I'm trying to target, then I might test differently. So if I know things bog down when a lot of people are commenting, I might design a JMeter test that hammers that. If I think it's about the sheer number of connections, I might use ab.
posted by advicepig at 8:28 AM on March 17, 2014 [1 favorite]
Of course, if there are specific things I'm trying to target, then I might test differently. So if I know things bog down when a lot of people are commenting, I might design a JMeter test that hammers that. If I think it's about the sheer number of connections, I might use ab.
posted by advicepig at 8:28 AM on March 17, 2014 [1 favorite]
Response by poster: Thanks for the answers. The site in question requires first loading a page, then clicking a button on said page, then waiting for the response (there's a hidden hash in the form, which is why we need to first load the page and then send the form). Would any of these tools be able to handle this workflow¿
posted by signal at 2:06 PM on March 18, 2014
posted by signal at 2:06 PM on March 18, 2014
« Older County Government Application Test | "I love you" 1) What does it mean? 2) When do you... Newer »
This thread is closed to new comments.
posted by holgate at 8:48 PM on March 14, 2014