Hybrid Orchestration
Hybrid orchestration runs game servers across both owned or reserved bare metal and rented cloud capacity, under one control plane. Baseline demand goes to the cheaper fixed capacity; peaks and unexpected regions spill into on-demand cloud. It is how studios reach bare-metal economics without bare-metal inflexibility.
Also called
hybrid cloud
,
How Does Hybrid Orchestration Decide Where a Match Runs?
A hybrid setup has two pools of capacity: fixed hosts, paid by the month, and cloud servers, paid while they run. The orchestrator sees both, and every match request goes through one rule. Start the server on a fixed host if one has room, and in the cloud if none does.
That rule works because the server is the same on both sides. The game server is packaged once as a container image, so it starts the same way on a reserved host or a cloud VM. The matchmaker asks for a server and gets an address back. Neither it nor the players know which pool the server came from.
A second rule decides the harder case: the fixed hosts have room, but they are far from the match's players. Filling the fixed hosts first keeps cost down, because that capacity is already paid for. Placing the match in a cloud location closer to its players keeps latency down. Both answers are reasonable, and a studio can answer differently by region or by mode, such as cost-first for casual queues and player-first for ranked.
How Big Should the Fixed Baseline Be?
Big enough to stay busy. Plot concurrent players over a typical week and find the floor: the level the game rarely drops below, overnight and midweek. Fixed capacity sized to that floor runs full around the clock, which is where its lower price per hour pays off. Everything above the floor goes to the cloud.
The two ways to get it wrong cost differently. A baseline too small sends a few extra matches to the cloud at the on-demand rate. A baseline too large leaves hosts idle at full price, every hour, until the term ends. When the curve is uncertain, erring low is the cheaper mistake.
The floor also moves. At launch nobody knows where it sits, so the mix leans on cloud. Once the curve settles, the baseline can grow to meet it, and as an audience shrinks, the baseline shrinks with it at each renewal. Shorter terms cost more per month and let the baseline follow the curve more closely.
What Does One Control Plane Change?
Running fixed and cloud capacity side by side is possible without hybrid orchestration. It usually means two systems: a cluster on the bare metal, often Kubernetes with Agones, and a separate setup in the cloud. Each has its own deployment pipeline, monitoring and placement logic, plus glue code to decide which one a match goes to. Hybrid orchestration replaces that with one API, one image and one view of every server.
One control plane also covers for a failing pool. If a provider or region goes down, matches can start in the other pool instead of not starting at all, the case our analysis of hyperscaler outages made after the November 2020 AWS outage (December 2021).
On Edgegap, the two pools are Private Fleet, reserved hosts in 58 locations, and Edge Cloud, 615+ locations across 17+ providers (platform data, 18 September 2026). A studio lists its fleet hosts in priority order. A deployment starts on the first host with room and, when none has, overflows to Edge Cloud automatically, at the best available location for its players at request time (private fleet documentation, October 2026). The priority list is where a studio sets its cost-first or player-first rule.
Hybrid Orchestration vs Multi-Cloud: What's the Difference?
They answer different questions. Multi-cloud is about providers: running on more than one cloud, to reach more locations or to avoid depending on one. Hybrid is about how capacity is bought: fixed hosts on a commitment, alongside capacity paid by use.
They combine, and for game servers they often do: a fixed baseline with one provider, cloud capacity spread across several. The difficulty grows with every provider added, because each has its own API, machine images, networking and billing. Our analysis of multi-cloud management put the limit of infrastructure-as-code tools at around two or three providers and 40 to 50 locations; beyond that, managing them by hand stops being practical (December 2022). That is the work a hybrid, multi-provider orchestrator takes on.
A word from our sponsor (ourselves!)
Hosting on a single cloud means you go down whenever that cloud does. Edgegap orchestrates your servers across 17+ cloud and bare-metal providers with automatic failover, so no single provider's outage takes your game offline.
Edgegap's Take (just our opinion, take it with a grain of salt!)
Hybrid Is a Policy, Not a Purchase
Studios usually come to hybrid as a buying decision: how many hosts to reserve. The hosts are the easy part. Whether hybrid saves money and keeps matches close to players depends on the policy around them: which pool a match goes to first, where it goes when the fleet is far from its players, and how often the baseline is resized.
The first two are placement rules, and they are where cost and player experience meet. A fleet filled first is cheaper. A match placed near its players plays better. We let studios set that order through their host priority list rather than deciding it for them, because a ranked queue and a casual one can reasonably answer it differently.
The third is a calendar. A baseline sized at launch is wrong a year later, in one direction or the other. Private Fleet terms run monthly, 6 or 12 months (bare metal hosting, October 2026), so the baseline can be resized at each renewal instead of once.
Set the policy first. The size of the fleet follows from the floor it reveals, and the pricing calculator shows what that mix costs at your own concurrency.
,
