Scalable Image Processing Using Hosted Cloud Services

blog.iron.io

17 points·kenfromm··7 comments

I just use chunky PNG for my image processing needs. ImageMagick is too bloated and contains features I don't really need. I call chunky to resize avatars and pictures uploaded once, and then use a delayed job to resize them into other formats that won't be needed immediately.

What would be the advantage of using this vs using Resque + S3 + heroku worker auto-scaling?

I've set up this exact flow using delayed_job, heroku, S3 and paperclip (an awesome gem wrapping imagemagick). I'm with jblake, I don't see the advantage.

Presumably: cheaper, easier set up. Sounds a lot like PiCloud.com

The way I understand their pricing (.05/hr, .10/hr, .15/hr based on minutes jobs get pulled off queue), I'm essentially paying 3x more for priority queuing that I can bypass by setting the INTERVAL=3 in Resque. Hm. This an expensive way to set a variable.

IronWorker is billed by the second so if an image takes 5 seconds to process, you only get billed for 5 seconds of the hour. No need to deal with auto-scaling heroku workers to keep your costs down.

And priority on IronWorker has a much different meaning than setting it on your own private Resque instance so comparing that based on price isn't really a fair comparison.

No resque setup/management, no auto-scaling setup, scale huge with zero effort (go from processing one image to millions without lifting a finger), redundancy and reliability (how are you going to ensure that with your Resque?), full visibility into your tasks, etc.

Reply on news.ycombinator.com