from Hacker News

How We Are Improving Performance

by snackai on 3/18/17, 3:20 PM with 55 comments

  • by gingerlime on 3/18/17, 4:23 PM

    We were using imgix for a while and were generally happy, but things started go downhill and some point, or so it felt anyway. Their support was always a bit opaque. And the service itself didn't evolve (as far as rendering. E.g. Composing more than one of the same filter wasn't really possible... for example adding two watermarks). We've also had issues with CORS headers that weren't resolved and our end users couldn't get images some times...

    We switched to host our own thumbor (open source), and couldn't be happier. We pay around a quarter or less than before (even with failover in place) as well.

    We really wanted to use a hosted service. We're not keen on hosting stuff out of our core business. But in this case it just didn't work out.

    EDIT: link to Thumbor https://github.com/thumbor/thumbor

  • by mabbo on 3/18/17, 3:38 PM

    There's two kinds of apologies, and they look very similar on the surface. There's "Sorry for this problem, but here's why it's not my fault" and there's "Sorry for this problem. Here's what happened". Am I taking the blame for the failure of others, or deflecting the blame onto others?

    This feels a lot more like the latter, and it's wonderful to see. I'm sure it's a lot harder to write, but it shows a heck of a lot more integrity.

  • by ShirsenduK on 3/18/17, 3:46 PM

    They put out a blogpost while their support keeps stonewalling and denying any such issues. Nor do they reply on twitter or even tweet the blogpost. This seems too little too late.

    Disclaimer: I am currently a customer who had financial loss because of them.

  • by trevyn on 3/18/17, 4:10 PM

    I wish this post gave more concrete details, so we could learn from it.

    A critical piece of our network infrastructure failed after 3 years of correct operation in a way that proved difficult for our network engineering team to troubleshoot.

    Which piece of infrastructure? How did it fail? Why was it difficult to troubleshoot? Can you do anything to prevent this type of issue in the future?

    We observed new traffic patterns with significantly lower cache hit rates than our historical median, and it took us some time to determine whether the source was abusive in nature or a legitimate new customer use case.

    What were the new traffic patterns, and why did this cause lower cache hit rates? Why did it take longer than expected to determine the nature of the traffic?

  • by rcchen on 3/18/17, 3:55 PM

    I wonder if their render service is still backed by racked Mac Pros (http://photos.imgix.com/racking-mac-pros). If so, considering the lack of updates around that machine for the last several years, I wonder if they are planning to remain with that solution.

  • by Nabi on 3/18/17, 4:02 PM

    Thanks for explaining and openness. We were bit frustrated as just finished migration to imgix and hit many issues with images delivery.

  • by mnutt on 3/18/17, 4:10 PM

    Improved our GIF encoding pathway to increase throughput.

    I'd be curious to hear exactly what they did around this. I recently worked on some gif encoding, and was surprised that there's actually quite a tradeoff to be made between performance and good color palette choices.

  • by microcolonel on 3/18/17, 3:53 PM

    The company that literally racks Mac Pro trashcans sideways, is telling us that they have made technical decisions that allow them to offer a competitive price for features.

  • by amelius on 3/18/17, 4:45 PM

    It says on their homepage that they use a 3rdparty CDN. I'm wondering how you can improve upon that if you are just a customer from the CDN's point of view.

  • by mkup on 3/18/17, 6:06 PM

    Font on https://www.imgix.com/pricing page is broken in Windows/Firefox (look how small "e" and "a" are displayed): https://image.ibb.co/ei7m8v/windows_font_problem.png

    This is without zooming (100%), 120 DPI, Windows 7 x64.

    Please use standard fonts and don't overengineer, font hacks like this will never work as expected across all platforms.

  • by Thaxll on 3/18/17, 4:28 PM

    Imgix is a good example of burden of legacy code / infrastructure that they can't get away from.