Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Is there a way that I can objectively compare the performance of different CDNs? Will I need to purchase a plan and test it out myself?

share

migrated from stackoverflow.com Oct 25 '09 at 23:52

This question came from our site for professional and enthusiast programmers.

locked by Mark Henderson Jan 22 '12 at 7:06

This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here. This question and its answers are frozen and cannot be changed. More info: help center.

    
Most companies let you set up a short term trial for free. –  Joel K Jan 22 '12 at 5:09

2 Answers 2

Here is an article that compares the latency between a number of different CDNs. Their results may be a little out of date, but you can redo the tests if you`d like using the same ping website they did.

share

A good way to objectively test a CDN is by using a monitoring service that's geographically diverse and on diverse networks that can mimic your application loads.

A good CDN will get content to your users at lower latency and with fewer hops than going all that way back to your origin servers.

For a web site, that typically means web page loads.

You didn't specify your application, so I'm assuming web site.

Companies like Gomez and Keynote (there are many others -- see some below, but those are the old time heavyweights-- most $$$$) offer services that track how well a webpage will load under a variety of parameters. For US focused sites, I usually like to check from 6 regions in the US. (Seattle, LA, Chicago, Dallas, Boston, Miami).. If you care a lot about international traffic you can set up probes abroad too.

You also want to setup probes on different ISPs. How well a CDN works for customers with Comcast might be very different from how it looks to customers on Level 3. Having a mix of top tier and 'eyeball' networks is good. (monitoring companies already have probes on a variety of networks)

Be careful with these services. They tend to charge per probe you hit, and while you may think that checking every 5 minutes is cool, it will soon cost you. Consider it a development tool to track long term trends and only turn up the frequency when you're A/B testing CDNs. (better yet, set up a different URLs for each CDN you are testing and let the same probes hits all of them.)

If you have some Javascript-FU, you can bypass a lot of what these services offer and set up your pages to report back to you about their load characteristics from user's browsers. (best to do a sampling, not EVERY page load) Couple that with a GeoIP database, and you can get some good baseline data to start with.. Here's someone doing that... HTML5 method for page load measuring

Cheaper alternatives to Gomez and Keynote: Pingdom WebpageTest Siteuptime AlertSite

I've had good luck with AlertSite in the recent past.

Cheers,

Joel

share