45

Source document

[2026-02-01 23:19] xenonfun: well its used heavily in game stats for Xbox stuff, fintech for trading things, IOT. if you need millions of active grains out of a set of billions, that is upper bound of scale they are trying to address, but you can run same stuff single silo cluster with like 20-30mb of overhead: https://learn.microsoft.com/en-us/dotnet/orleans/ they updated there docs just recently, less confusing as they had a lot of Orleans 3.0 to 7.0 hard migration notes, its at 10.0 now. [2026-02-01 23:19] xenonfun: but think most large deployments are kubernetes or service fabric on Azure [2026-02-01 23:20] traves_theberge: right so why Orleans? [2026-02-01 23:21] xenonfun: cause I am in .NET and its native virtual actor setup, other options would be Akka.net think there is one other actor setup [2026-02-01 23:22] xenonfun: https://aka.ms/orleans-discord they have discord that seems kinda active especially for real production issues and such people see [2026-02-01 23:23] xenonfun: they also are more opinionated in framework, which means its way less to deal with than frameworks with more freedom. [2026-02-01 23:24] xenonfun: Erlang / Elixer would be other logical choice for building something you want to scale this way. [2026-02-01 23:26] traves_theberge: yeah makes sense, but practically, what is taking place in these cointers and clusters or what could one do. ill check our the docs now. [2026-02-01 23:27] xenonfun: well you can run silo whereever, to form cluster it would have to have network to talk to others. practially you would want some sort of container setup you spin up new instances with load as opposed to provisining servers. but they have a lot of options. because you might actualy want to segerate certain daemons to only host some grains in them, etc. [2026-02-01 23:28] traves_theberge: yes. [2026-02-01 23:28] traves_theberge: absolutely yes [2026-02-01 23:28] xenonfun: that way you can just upgrade certain grains at a time, its designed to not bust Halo game stats [2026-02-01 23:28] xenonfun: and still add features players want when they get built out [2026-02-01 23:29] traves_theberge: That would be nice to reduce the "FUCK VERCEL" rants im going on about. Railway seems like parallel with this idea. [2026-02-01 23:35] lisamegawatts: coolify is an alternative you can self host, railway, render, and heroku i think are parallels [2026-02-02 00:34] xenonfun: (files: Screenshot_2026-02-01_at_7.34.25_PM.png) [2026-02-02 01:22] xenonfun: adding Seq for log injestion and filtering. gotta be able to easily work with production produced exceptions: https://datalust.co/ put it into production at ArcBest it was really useful to capture all sorts of stuff. love its concept of a loghash (when using structured logging the template becomes a type so you can filter them better). Adding to docker container setup for local dev as well. [2026-02-02 01:45] xenonfun: (files: Screenshot_2026-02-01_at_8.45.42_PM.png)