Platform

Allocate

Matchmaker calls allocate (region, fleet, build config) - Multiplay spins up one game server process on a best-fit host.

Allocations

Allocator returns IP + port. Matchmaker hands connection details back to the client. Players connect directly to the game server. Match runs.

Deallocate

Session ends process exits, or matchmaker calls deallocate. Allocator releases capacity.

| Cost efficiency |

You pay for what you run

Hosting cost depends on the mix. All cloud means paying elastic rates for capacity that never changes. All bare metal means owning hardware sized for a peak you rarely hit. The sweet spot sits between them, where each session runs where it costs least.

Every server runs on rented cloud, the steady-state baseline included. You pay elastic rates every hour of the year for flexibility you only need at the peaks, which is why all-cloud sits at the top of the curve.

Steady-state runs on owned bare metal at a predictable cost, and cloud covers only the bursts on top. Each session runs where it costs least, so the total lands at the bottom of the curve. This is where most games sit best.

Every server runs on owned hardware, so there is no cloud premium. The trade is that you own enough capacity for your highest peak, and most of the year a lot of it sits idle. Lower than all-cloud, still above the sweet spot.

Illustrative annual hosting cost across the bare-metal / cloud mix.

Backend provisioning

VM, container, and bare metal provisioning, with OS, network, and security policies applied.

Security

Identity, secrets, environment-scoped authorization, a patching cadence, and a compliance posture.

Build, deployment, patching

Build configs, mirror and CDN delivery, OS and game patching, build promotion across environments.

Live operations monitoring

Fleet health, drift detection, and alerting. The platform kept healthy on your behalf.

Observability

Logs, crash reporting, and tracing, tagged to every allocation, and customer-facing.

Hybrid scaling

The autoscaler. Pre-warmed proactive scaling plus reactive burst.

Capacity planning

Sizing for your title, the bare metal and cloud mix, and launch ramp planning.

Integration

SDK and matchmaker compatibility. The SDK is optional, and direct API is supported.

Architectural support and TAM

A named technical account manager per tier, co-engineering at the architecture level.

| GLOBAL FOOTPRINT |

WHERE WE HOST

30-plus locations across six regions, on owned bare metal and multi-cloud, with redundancy at the region level. A sub-30ms target across major population centers.

NORTH AMERICA · 16 locations

AshburnAtlantaChicagoDallasLAMiamiNYCNewarkPortlandSalt LakeSan JoseSanta ClaraSeattleSt LouisTampaWashington DC

Europe · 9 locations

AmsterdamFrankfurtIstanbulKyivLondonLuxembourgMilanRotterdamStrasbourg

Asia · 7 locations

MumbaiNew DelhiSingaporeHong KongSeoulTaipeiTokyo

LATAM · 3 locations

Buenos AiresSantiagoSao Paulo

MEA · 3 locations

JohannesburgTel AvivDubai

Oceania · 2 locations

SydneyMelbourne

| THE PLATFORM |

WE HAVE THE PLATFORM THAT SET THE STANDARD

| HYBRID ARCHITECTURE |

We blend metal and cloud

| HOW ALLOCATION WORKS |

We do the hard part, just tell us when you need capacity