These type of articles are always fun, though I mistakenly thought they meant Discord was running on a single hardware server, that would be fun to read. Though considering they run on BEAM they can spread out the workload plenty. Still, I remember when the hard limit for a Discord used to be 1 million, and then along came MidJourney.
It's funny how in hindsight simple changes to major web applications have a major impact. I remember, Tumblr used to show you every statistic about your blog on the dashboard, that meant every single time someone hit the front-page they had to query the latest data to show you your follower count. Sometime after they removed it from the Dashboard, I noticed Tumblr went offline / down significantly less, I doubt that was the only optimization they did, but I presume they did loads of optimizations to their platform at the time.
Yeah, statistical things like that need to be pre-calculated and cached for best performance.
When you first get stuff working doing a query is easier and faster. At some point you remove it and move to counters that are updated via your data pipe.
It is important to node here that by "Server" they mean the Discord logical server concept which may more generically be called a "community", "group" or "workspace". It has nothing to do with the physical compute architecture.
So a title that would make sense to people not familiar with Discord terminology could be:
> How Discord Serves 15M Members of a Single Workspace
8 comments
[ 3.2 ms ] story [ 31.1 ms ] threadIt's funny how in hindsight simple changes to major web applications have a major impact. I remember, Tumblr used to show you every statistic about your blog on the dashboard, that meant every single time someone hit the front-page they had to query the latest data to show you your follower count. Sometime after they removed it from the Dashboard, I noticed Tumblr went offline / down significantly less, I doubt that was the only optimization they did, but I presume they did loads of optimizations to their platform at the time.
When you first get stuff working doing a query is easier and faster. At some point you remove it and move to counters that are updated via your data pipe.
So a title that would make sense to people not familiar with Discord terminology could be:
> How Discord Serves 15M Members of a Single Workspace
[0] https://discord.com/blog/maxjourney-pushing-discords-limits-...