Differences

This shows you the differences between two versions of the page.

--- commissies:sovcom:start [2026/02/05 12:40] – [Local compute cloud] haaijer
+++ commissies:sovcom:start [2026/03/19 13:36] (current) – [Local office suite] haaijer
@@ Line 2: / Line 2: @@
 This is not an official committee (yet). If you are interested, please contact haaijer@snt.
+This document is never done. Please extend it with new learnings and choices as time goes on.
 SNT has always been a big proponent of open source and self-hosting. Given the current (~2025) happenings in the US, decision makers in Europe and the UT are also starting to share these ideas.
@@ Line 16: / Line 18: @@
     * Missing some nice to haves such as reinstalls, cloudinit scripts
   * Cloudstack
-    * Less popular but seems to be growing
+    * Maintained but not extremely active
+    * Becoming more popular now
     * More cloud-based, so end users can make their own VMs on demand
-    * We need to look into this more.
+    * Requires Alma/Rocky
+    * A bit overkill for SNT, but great for organisations that want more features and a better UI compared to proxmox
   * Openstack
     * By far the most features
@@ Line 42: / Line 46: @@
 ===== Local chat =====
 TODO: Matrix, Mattermost, Discord, Simplex, UT situation, etc
+Within the University of Twente, many technical studies used to have a Discord group for every module for students to communicate with eachother and with teachers for questions. This was a great way to get faster support from peers and enabled a searchable and accessible platform.
+Since end of 2025, the UT has decided to ban the use of Discord. This is not completely unfair since it's an American platform likely owned by Tencent (chinese state). Unfortunately, no good alternative was given.
+One department was already running mattermost and was happy with that. Unfortunately, due to licensing this is not an easy option. Most departments switched to teams, but this is has worse UX and is still Microsoft.
+SNT has been running a large matrix server for a few years. First publicly and then later limited to Dutch universities. While there are some technical problems with the main server implementation Synapse, the UI/UX is described as "Not worse than M$ teams".
+SNT is currently evaluating options for onboarding other associations, and perhaps even helping the UT set up an instance as the main chat platform for student communication.
 ===== Local office suite =====
-TODO: Opendesk, LaSuite, MijnBureau
+WIP.
+It would be nice to offer an office suite to associations that is not google or microsoft, and perhaps roll that out to the full UT.
+Some popular options:
+  * Nextcloud
+    * Mostly for file management, but rich plugin support
+    * UT (Henri) is working on this through SURF.
+      * Federated auth (so external uni employees can collaborate)
+      * UT plans support for UT 053 phone numbers
+    * quite well integrated
+    * Plugin system means that some development is a bit harder, but its viable
+  * Opendesk
+    * Most complete.
+    * German government project
+    * COllection of pre-existing software. Mainly nextcloud, openproject, matrix, collabora
+    * If you want to setup a testing environment on a local machine, see [[https://wqrld.net/blog/setting-up-opendesk-on-kubernetes/|https://wqrld.net/blog/setting-up-opendesk-on-kubernetes/]]
+  * LaSuite
+    * Some pre-existing software but also a lot of custom programs
+    * Personal opinion: Better UI
+    * No official deployment options
+  * MijnBureau
+    * Dutch government
+    * Combination of LaSuite and Opendesk components
+    * Looks quite promising
+    * Still very much work in progress, but is starting to be usable due to some work by an SNT'er
 ===== Local LLM =====
-TODO: snt, UT/EEMCS
+SNT has a workstation GPU available in the room for anyone wanting to experiment with running local LLM's. It can run deepseek etc. at decent speeds.
+The easiest way to get started is with Ollama + [[https://docs.openwebui.com|Open-webui]]. Keep in mind that running without a GPU is really really slow.
+Local LLM's do not yet have the quality that the big commercial ones do, partially due to hardware limits and partially due to the models. They do however get quite close.
+The UT (LISA+EEMCS) are working on procuring some GPU's to set up a local LLM for UT students and researchers.
+If necessary it might also be possible to lend some unused compute from one of the research groups?
 ====== Interested in working on this? ======