How it works
Server requirements depend mainly on how many learners are logged in at the same time, and on what they are viewing. The simulator starts from the number of enrolled learners (from 500 to 50,000) and a share of logged-in users (5 to 15%), then estimates the processor, memory and bandwidth required, and the architecture to plan for.
The simulator
Choose your assumptions: the charts, the recommended configuration and the benchmarks update.
How to read the results
- Processor and memory: rounded up to the next server size, with at least 2 GB per vCPU, as hosting providers do. The installed memory is therefore often higher than the memory used shown in the chart.
- Outbound bandwidth: this is the first limit you hit when courses contain a lot of video. Beyond a 1 Gbit/s link, a CDN (content delivery network) takes over for videos.
- Architecture: beyond 16 vCPU or 64 GB, a single machine becomes a single point of failure; the database is then separated from the web servers, behind a load balancer.
The assumptions
- A typical learning session lasts 30 minutes: one visit to the dashboard, two to "My courses", two modules opened, and progress saved about once a minute.
- A module delivered on a partner's site (Wrapper) loads a dozen files per minute, each one checked by the platform.
- Videos tie up the server while they download: they are the first factor for memory and bandwidth.
- Load peaks are taken into account (×1.5 on the processor, ×2 on simultaneous connections). The database accounts for about a third of the load.
- "Over the day": a 35-minute session, with 20% of the day's logins concentrated in the peak hour.
The production minimum is 4 vCPU / 8 GB. Not included: large exports and reports on the administration side, or a locally hosted AI assistant (in that case, allow +4 vCPU and +8 GB). These values are estimates to help you choose an order of magnitude; beyond 10,000 enrolled learners, we recommend a load test on your environment.