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

We have been running a website that contains quite large image gallery. We started growing and we need a new place to keep our images. So I started looking and testing.

For testing purposes I created three links that are identical (200 divs with images in them), apart from the links they are loading images from.

So we uploaded 200 images to RackSpace CloudFiles, RackSpace VM with NGINX and our current server in Iomart UK.

I was using Loads.in tool to check the loading time of each link from all the viable locations. These are the results: (Tested mainly Europe as thats where we got most clients)

Todays results:

screen-shot of results

Week old results:

screen-shot of OLD results

In the old results it was Clear - CloudFiles CDN is slower then just a plain server on Iomart with NGINX. But the new results (1st link) is kind of mixed results and not really that clear.


I have also done some tests using curl:

curl http://foo.bar/image.jpg -o /dev/null

I run the same command for 100 times and then calculate the average speed, and showed that that CloudFIle links were slower then Iomart server links.


In conclusion, the result really feels when you run it live. We have tried running gallery on CloudFiles, but you could clearly see how much slower it was. Now you could say that the results were different because of my location, but as long as curl results were pretty much the same from my location, germany and UK london.

Now after few weeks of trying to find a good place to fit our gallery I am getting more and more confused and started questioning my methods.

As much as I believe that CloudFile with Akamai, should be faster - according to my tests - it is not. And you can really feel how slower it is using and browsing the CloudFiles and Iomart galleries.

Am I doing my tests in a bad way ? Or these are normal results and I simply do not understand something ?

Has anyone had some similar experience and could give me an advice on this ?

share|improve this question
    
Remember that speed isn't the only possible benefit from a CDN. Saving you from getting overloaded with static file requests is another. –  ceejayoz Oct 1 '12 at 15:43
    
I do agree, but I need speed. I will find another solution to number of files. Speed is my priority at this point. –  Katafalkas Oct 1 '12 at 15:47
    
If I understand correcty, you execute curl once per image, so one tcp connection per http get request. Most browser don't behave like this : en.wikipedia.org/wiki/HTTP_persistent_connection . You may want to have a look at phantomjs, jmeter or neoload for testing. Also consider using varnish+nginx serverside. –  coincoin Oct 1 '12 at 15:50

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.