1. 7

  2. 5

    Seems like a step backward…

    1. 1

      I’m not the biggest fan of AWS because I think it’s too big and I don’t really want to give Amazon any of my money (not that that would make a difference to them…). But they do make things quite easy and I can imagine it’s less likely for CloudFront to go down than a single dedicated server.

    2. 2

      Running mediawiki on AWS is nearly impossible; if it is possible, it would be very expensive, an amount we wouldn’t be able to pay.

      I wonder why. As far as I know, MediaWiki is just a typical database-backed PHP application. It shouldn’t be difficult or expensive to run that on AWS.

      1. 1

        That really seems like a bizarre claim, yeah - particularly for infrequently-written workloads. A single, small, reserved instance running db+webserver should be ample; add a nightly database backup to s3 (& test restoring it on a new instance).

        There’s even a debian package; apt install apache mediawiki on an empty ubuntu image should get you nearly all the way there.

        1. 2

          MediaWiki’s parser is surprisingly CPU-intensive, especially once you start using templates. The whole Lua scripting improves on that a bit, but it’s still much more expensive then you’d think. I think at some point the Barrack Obama article took like a minute to render in full (without caching). The only reason it appears fast is because it caches so much.

          1. 2

            There’s some discussion here, but I’m not sure it really maps to “running mediawiki on AWS is impossible”: https://github.com/TrueBrain/TrueWiki#why-yet-another-wiki-server