OpenSimulator service architecture
Standalone, ROBUST, regions and Hypergrid: where should each process run?
Standalone: one combined process
Standalone combines simulator and required grid services in one deployment, ideal for learning and smaller setups. It does not require a separate ROBUST deployment. Database, directories and actual ports depend on your configuration.
Grid: separate ROBUST and region simulators
In Grid mode, ROBUST runs central services such as accounts, authentication, inventories, assets, grid and presence depending on the enabled modules. OpenSim runs regions, avatars, scripts and physics. One ROBUST deployment may serve several OpenSim processes on one or more authorized machines.
Why split ROBUST from regions?
Physics and scripts can consume CPU while SQL and assets may be more sensitive to storage and network behavior. Splitting workloads helps diagnosis, failure isolation and maintenance. However two VMs on one physical host still share hardware: logical separation is not redundancy.
Hybrid grid and external regions
A grid may host ROBUST at a provider while residents operate their own regions on separate hosts. Each simulator needs permitted access to central services, and viewers must reach its public region ports. This requires suitable firewalls, accurately advertised addresses and a trust policy between operators.
Hypergrid does not replace networking
Hypergrid enables compatible grids to exchange visitors and other interactions but does not remove public reachability, DNS, certificate or port requirements. Test both external and local visitors, not just a connection from the server itself.
Document a network flow map
For each component, record host, private IP, DNS, port, protocol, allowed clients, logs and backups. This traffic matrix makes handshake failures, SQL errors and firewall denies easier to diagnose. Internal services such as private ROBUST endpoints should not be open to arbitrary internet sources.