Server Allocation
Server allocation is the process of assigning a dedicated server instance to a specific match or session — the moment between a matchmaker finding players and those players actually connecting to a running game. The matchmaker decides who plays together; allocation decides where they play.
What happens during allocation
When a matchmaker fills a lobby, it sends an allocation request to an orchestrator. The orchestrator either starts a new server instance for that match or assigns an already-running idle instance. Once the server is ready, the orchestrator returns a connection address — IP and port — to the matchmaker, which hands it to the players’ clients so they can connect.
That handoff chain is: matchmaker → orchestrator → server start → connection address → players connect. Every step adds latency. Most players never see it, but if allocation is slow, they stare at a loading screen while the server warms up.
Pre-warmed vs. on-demand allocation
There are two fundamental allocation strategies:
On-demand allocation starts a fresh server instance only when a match is needed. Cost-efficient — you pay for nothing until it is needed — but startup time adds to player wait time. On cloud VMs, where containers take 30–90 seconds to start, this creates a noticeable delay.
Pre-warmed allocation keeps a pool of idle server instances running and ready to accept a match. Allocation completes in milliseconds because no startup is required — the server is already running. The tradeoff is paying for idle capacity. How large that pool needs to be depends on your expected match rate and the variance in demand.
Platforms that run on bare metal can start server instances in under a second, which effectively collapses the on-demand startup penalty and makes large pre-warm pools unnecessary.
Why allocation speed matters at launch
At launch, match requests arrive faster than they ever do during normal operation. If your orchestrator is allocating on-demand from scratch, a spike in demand means a spike in startup latency — players wait longer exactly when first impressions count most. A platform that can allocate in under a second absorbs the spike without players noticing.
Allocation and session data
A good allocation system returns more than just a connection address. It tracks which server is running which session, attaches metadata — match ID, region, game mode, player count — and monitors whether the session is healthy. That data feeds back into the orchestrator so it can release the server when the session ends rather than leaving it running idle.
See also: How Gameye handles session scaling · Matchmaker · Orchestrator · Session Hosting