System Requirements
This page lists what a host needs to run the whole InfoConnect system stably: the Hub and the optional RPA Hub, MCP, UI and Auth server, all on one machine. "Stable" means every module starts, none is killed for lack of memory, the CPU is not saturated for long, and data flows without errors. The databases of the modules, the target database and disk space are sized separately. For the software prerequisites, such as Java 17.0.1 or later, see Hub infrastructure.
| Minimum | Recommended | |
|---|---|---|
| CPU | 1 vCPU | 2 vCPU |
| RAM | 4 GiB | 8 GiB |
The minimum runs the system with a light data flow; the recommended size keeps it comfortable under a heavy one. More resources mean a faster data flow, up to the limits of the IBM i and the target database. The figures include the operating system and assume a typical load, a target database close to the Hub and the default settings.
- Hub heap. The start scripts limit the Hub heap to 1 GiB (
-Xmx1024min the[jvm.hub]section ofstart-scripts/config/modules.manifest.ini). To use the recommended size, raisememorythere. - Only the Hub grows with the data flow. The other modules keep a small, constant footprint: about 1.3 GiB of RAM together and almost no CPU.
- Start-up. Starting all modules at the same time takes about 2.7 minutes on 2 vCPU and about 4.8 minutes on 1 vCPU. On 1 vCPU keep the start-up timeout of a service manager or orchestrator at 5 minutes or more, or start the modules one after another.
- Separate hosts. If the modules run on different hosts, the host with the Hub needs about 1 GiB of RAM at the minimum and 4 GiB at the recommended size plus the operating system; the host with the other modules needs about 1.5 GiB of RAM and very little CPU.
- Other conditions. A target other than SQL Server (the only one measured), a much larger batch size, or a large distance to the IBM i or to the target needs more memory and CPU: start from the table and watch both.
Basis of the figures
The Hub was measured under load with SQL Server as the target, on Kubernetes (AKS) and on a virtual machine with 2 vCPU. The other modules were measured idle and at start-up, so the totals for the whole system are derived from these measurements and not measured with all modules running together under load.