Scaling and load balancing

Your deployments grow and shrink with the traffic

Traffic is shared across healthy containers, containers are added under load and removed when it drops, and new versions go live without dropping a request.

How MultiConnect Cloud serves trafficBrowsers and mobile apps reach website and API containers through an HTTPS edge that balances load. The API scales from two containers to three under load and reads from a database with automatic failover to a standby. Game clients connect straight to a game server that starts for one match and is removed when the match ends.BrowserMobile appGame clientHTTPS edge · load balancing · your domainown IP and portTCP, UDP, WebSocketWebsitestatic, HTTPSrollback keptAPI2 runningscaled to 3autoscaling on CPU and memoryGame servermastermatchstarted for this matchmatch over, billing stoppedDatabasemanagedprimarystandbyPostgreSQL or MySQL.Automatic failoverto a standby.

What you get

Included with API, website and server plans

Autoscaling on CPU and memory

Set a minimum and maximum container count and the thresholds that trigger scaling. For API, website and master-worker server deployments.

Load balancing

Each request goes to the next healthy container in turn. Included for APIs, websites and secure WebSocket servers.

Health-aware routing

A container that fails its health check is taken out of rotation and put back when it recovers.

Deploys without downtime

A new version starts next to the current one and receives traffic only once healthy. The previous version is kept for rollback.

Graceful server rollouts

For game and realtime servers, old servers stop taking new sessions and drain for a grace period so connected users are not cut off.

Limits set by the plan

Each plan sets how many containers you can run per region.

Autoscaling

How it decides

Measure

CPU and memory are sampled for every container.

Scale up

When average usage stays above your threshold for a few minutes, a container is added, up to your maximum.

Scale down

When usage stays below your lower threshold for longer, a container is removed, never below your minimum.

Stay stable

Scale-down waits longer than scale-up, so a short dip does not switch containers on and off.

Every running container is charged by the minute. Your maximum container count is also your cost ceiling, so set it with your budget in mind.

Questions

Does load balancing cost extra?

No. It is included. You pay for the containers you run.

How many containers can I run per region?

It depends on the plan; the dashboard shows the limit when you choose one.

Do databases autoscale?

No. A database has the size its plan sets. Change plan when you need more.

Do combined server deployments autoscale?

No. Combined mode runs one server. Use master and workers for a pool that scales, or on-demand servers for one per match.

Try it with 500 free credits

Create an account and you get 500 credits to deploy with. No credit card required. There is no free hosting tier; the credits are the trial.