Alpha tests
Alpha tests are the first round of game testing, run early in development to identify glitches, design problems, and technical issues.
Read more
Alpha tests are the first round of game testing, run early in development to identify glitches, design problems, and technical issues.
Read moreAnti-cheat software checks for wall hacks, aimbots, boosting, and even whether players are using scripts to cheat.
Read moreAn authoritative server runs your game logic and determines match truth — no client can override it, making cheating structurally impossible.
Read moreAuto scaling starts new game server instances as player load rises and releases them when demand drops — no manual intervention required.
Read moreYour backend is everything that’s invisible to the player. It’s all the tools and services behind the scenes.
Read moreBackfill adds new players to an ongoing match when others disconnect or drop out, keeping sessions full.
Read moreA bare-metal server is a physical machine which lives inside a data centre. It’s a place where you’ll host your games matches and sessions.
Read moreBeats is a family of lightweight Elastic Stack shippers that collect logs, metrics, and events from servers and forward them to Logstash or Elasticsearch.
Read moreBeta testing is when you get a group of players to test your game, typically when your game is close to launch day.
Read moreCCU (concurrent users) is the number of players connected to a game at once — the primary metric for sizing server infrastructure.
Read moreA Content Delivery Network is a network that hosts the same content in multiple places to get it to the end user faster.
Read moreClient-server architecture is where you have external machines that host your game’s server and players can connect to it.
Read moreClient-side prediction applies player inputs locally without waiting for the server, so movement feels instant even on high-latency connections.
Read moreCloud bursting is when game server workloads automatically spill from dedicated bare metal into cloud providers during demand spikes.
Read moreYour cloud server is a virtual server. It runs in a cloud computing environment, and it's a place where you’ll host your games’ matches and sessions.
Read moreContainerisation packages your game server, dependencies, and config into a portable image that runs consistently across local, cloud, or bare-metal infra.
Read moreA container is a packaged, portable software unit. For game servers, it’s your entire server build — easy to deploy, replicate, and roll back.
Read moreContinuous integration is when you regularly merge changes to your code from multiple people into one version.
Read moreA data center is a facility that houses servers used to store, process, and manage large volumes of data for applications and games.
Read moreA data packet is game data — position, inputs, events — sent across the network. Packet size and send rate directly affect latency and bandwidth use.
Read moreA Distributed Denial-of-Service (DDos) attack is where criminals attempt to overload your servers and cause them to crash.
Read moreA dedicated server is a remote machine that runs your game logic independently of any player's client — the authoritative source of truth for every match.
Read moreA dedicated server runs on separate hardware from any player. A listen server runs on one player's machine, making that player both host and participant.
Read moreDeployment releases a game server build to live infrastructure — containerising it, pushing to a registry, and starting instances so players can connect.
Read moreDesync occurs when the game state on a player's client diverges from the server's authoritative state, causing gameplay inconsistencies like phantom hits or invisible players.
Read moreDocker is a specific piece of software that combines your game, libraries, and config files into a single container.
Read moreDowntime is when a machine, or computer, crashes or is out of action. It’s rare, but can happen, especially during peak times.
Read moreAn edge node is a physical or virtual machine that acts as a bridge between a local network and the internet.
Read moreEgress fees are charges cloud providers apply to data leaving their network. For multiplayer games, egress typically represents 40-60% of total infrastructure cost.
Read moreThe Elastic Stack (ELK) is an open-source suite for ingesting, storing, and visualizing operational data, built from Elasticsearch, Logstash, Kibana, and Beats.
Read moreElasticsearch is a distributed search and analytics engine for indexing, storing, and querying game server logs, telemetry, and operational data at speed.
Read moreFailover switches to a backup system when a primary server fails — keeping your game online for players without manual intervention from your team.
Read moreFleet management is the process of provisioning, monitoring, scaling, and maintaining a pool of game servers across regions to meet player demand.
Read moreFlex metal servers are basically bare-metal machines, but you can have them available in the same amount of time as cloud servers.
Read moreGame server hosting is the service of running dedicated game servers on managed infrastructure so studios don't have to provision, scale, or maintain their own hardware.
Read moreGame server orchestration automates server management — session scheduling, scaling, placement, health monitoring, and failover via a single control layer.
Read moreA headless server runs without graphics, display, or audio — only game logic and networking — making it far cheaper and faster to host at scale.
Read moreA hitbox is the invisible collision shape on a character that determines whether an attack or bullet registers as a hit.
Read moreHost migration is the process of transferring a game session from one server to another mid-match, typically when the original host disconnects or fails.
Read moreIdle machines are provisioned server instances not running an active session. A warm pool of them reduces game server start-up latency for players.
Read moreA game server image is the packaged, executable version of your dedicated server code — built once, deployable to any container-compatible infrastructure.
Read moreGame server infrastructure is the hardware and software that hosts dedicated servers and keeps multiplayer sessions running reliably at scale.
Read moreInfrastructure automation provisions, configures, and scales servers using code — reducing human error and speeding up deployments without manual steps.
Read moreInterpolation smoothly renders other players between server updates so characters move fluidly rather than teleporting each tick.
Read moreJitter is variation in latency over time — inconsistent packet arrival that breaks interpolation and makes games feel erratic.
Read moreKibana is the Elastic Stack's visualization layer — build dashboards, explore logs, and monitor game server metrics stored in Elasticsearch in real time.
Read moreLag compensation rewinds server game state to the moment a player fired, judging shots fairly despite network delay.
Read moreLatency is how fast data is moving. How quickly does it get from point A to point B? (And it’s generally measured in milliseconds.)
Read moreLoad balancing distributes player connections across servers so no single machine becomes a bottleneck — keeping your game responsive under heavy traffic.
Read moreA lobby is the pre-match space where players gather before a game session starts — between matchmaking and gameplay.
Read moreLogstash ingests logs and events from multiple sources, normalizes them, and forwards them to Elasticsearch for storage and analysis.
Read moreA matchmaker groups players into sessions by running player data — location, skill, latency — through a set of rules to find the best possible match.
Read moreA microservice handles one function — matchmaking, auth, or session management — as an independently deployable component in a larger game backend.
Read moreMMR is the hidden score that matchmaking systems use to measure player ability and find balanced opponents.
Read moreMonitoring tracks known metrics — CPU, memory, session counts, latency — and alerts on threshold breaches so teams catch problems before players notice.
Read moreA multi-provider architecture distributes game servers across multiple independent infrastructure providers in each region, eliminating single-provider dependency.
Read moreNetcode synchronises game state between players. Poor netcode causes lag, rubber-banding, and hit registration issues in multiplayer games.
Read moreNetwork egress is what cloud providers charge for outbound data — the largest hidden infrastructure cost for game studios, and the most underestimated.
Read moreThe noisy neighbour problem occurs when one tenant on a shared cloud server consumes excessive CPU, memory, or I/O, degrading performance for other tenants on the same hardware.
Read moreObservability is understanding your system's internal state from its outputs — logs, metrics, traces — answering not just that something failed, but why.
Read moreAn orchestrator automates game server lifecycle — starting sessions on demand, placing them in the optimal region, and shutting them down when matches end.
Read morePacket loss is when network data never arrives. Even 1-3% loss causes rubber-banding, missing inputs, and broken hit registration in games.
Read morePeer-to-peer designates one player as host; others connect to their machine. Client-server uses external machines to host game sessions for all players.
Read morePing is a player's round-trip time to a game server in milliseconds. Lower is better — above 100ms feels noticeable in fast-paced games.
Read morePlayer concurrency is the number of players connected to game servers at the same time. It determines how much server capacity is needed at any given moment.
Read morePort mapping connects a container's internal ports to external ports on the host machine, allowing players to connect to game servers running inside containers.
Read moreA relay server forwards traffic between players without game logic — it solves NAT traversal but does not make your game authoritative or cheat-resistant.
Read moreA code repository is the version-controlled store for your game's source and assets — track changes, manage branches, and trigger deploys from one place.
Read moreReserved capacity is a committed block of compute resources at a discounted rate, used as a cost-efficient baseline for predictable game server workloads.
Read moreA resource profile will figure out how many resources your game will take, and in turn, how much compute power a server needs to reserve to run your game.
Read moreRollback netcode lets each client simulate inputs instantly, then roll back and resimulate when predictions are wrong.
Read moreRubberbanding is when a player's character appears to snap back to a previous position, caused by the server correcting the client's predicted state.
Read moreA sandbox is an isolated non-production environment for testing game server builds and integrations without affecting live players or production costs.
Read moreScalability is infrastructure’s ability to handle rising player demand — more sessions, regions, and traffic — without degrading performance or manual ops.
Read moreServer allocation assigns a dedicated server to a match — the step between matchmaking and players connecting to a live session.
Read moreA server browser is an in-game UI that lets players manually select which game server to join, as opposed to automatic matchmaking.
Read moreA server fleet is the pool of game server instances — running or on standby — managed by an orchestration platform, scaling up and down with player demand.
Read moreServer regions are geographic locations where game servers run. Placing servers near players is the most effective way to reduce multiplayer latency.
Read moreA serverless game server spins up on demand and terminates when the session ends — pay only for active play time, with no idle servers sitting unused.
Read moreSession hosting is when a server, or computer, runs your game’s matches. Each match, with a set group of people, is considered a session.
Read moreSession-based multiplayer is a game architecture where matches have a defined start and end. A server is allocated when the match begins and released when it ends.
Read moreSkill-based matchmaking (SBMM) groups players by measured ability so matches are competitive, not lopsided.
Read moreAn SLA is a contractual guarantee of platform uptime and performance. Game server SLAs typically guarantee 99.9% to 99.99% availability.
Read moreSoft launching is when you release your game to a specific market or a restricted audience, before your full launch.
Read moreStaging is the final pre-launch environment where your game runs in production-like conditions to validate server builds before real players connect.
Read moreTick rate is how often a game refreshes its information. Usually, it’s measured in hertz. For example, a 128-tick server will update 128 times a second.
Read moreUDP vs TCP for multiplayer games: TCP guarantees delivery and order; UDP trades reliability for speed. Most real-time games choose UDP.
Read moreA vCPU (virtual CPU) is a unit of compute capacity representing one logical processor core. Game server hosting is typically priced per vCPU-hour.
Read moreWarm pools are pre-started containers held in reserve so game servers can be assigned to players instantly instead of cold-starting on demand.
Read moreZero-SDK architecture is a game server hosting approach where no SDK, plugin, or code changes are required inside the game server binary. The platform manages everything externally.
Read more