Backends that survive a launch weekend
Game server infrastructure fails in a specific way: everything works in soak tests, then launch day multiplies traffic and the weakest dependency folds. Appliscale is a game backend development company with Erlang/OTP as a native competency and 11+ years of distributed systems work behind it.
Our engineers embed in existing game teams rather than bolting on a parallel org. We take services from design document to production deployment, monitoring and on-call, and we keep client identities confidential by default. For the full gametech practice, see gametech.appliscale.io.
Messaging and social systems at 100 million players
For one client we modernised the chat and notification platforms used by multiple game teams. The work included upgrading Erlang service runtimes, migrating storage from Riak to Redis and Aurora, and moving deployments from raw EC2 to Kubernetes.
The result: systems that scale to 100 million monthly active users and up to 4 million concurrent players, with an estimated 50K USD per month cut from the AWS bill and fewer incidents pulling engineers off feature work. Read the details in the Erlang messaging services case study.
Player identity and authentication at scale
Identity is where game backends concentrate risk: one slow login service degrades every title on the platform. We run a cluster of identity work for major gaming platforms.
- Authorization server. A phased migration from a debt-laden TypeScript service to Kotlin and Spring Boot, now handling 300K+ requests per minute at peak, hundreds of millions per day, with vulnerable dependencies removed along the way. See the gaming authorization server case study.
- Login platform. Services for 15M+ daily active users, migrated to Spring Boot 3.2 and Java 21. A Redis caching layer cut S3 costs about 40 percent, load-test-driven downscaling cut operational costs about 30 percent, and monthly invoices dropped 40 percent. Details in the login application case study.
- Unified MFA. An API-first MFA Factor API supporting mobile push approvals, TOTP and third-party authenticators, delivered by engineers embedded across backend and client SDK teams so it works consistently on PC, mobile and console. See the multi-factor authentication case study.
This is the kind of work engineering leaders bring us in for; if that is you, see how we work with CTOs.
Live ops infrastructure and real-time game data
Live ops runs on data paths that cannot lag. We built a game-agnostic leaderboard platform on Aurora MySQL, Redis and Kafka that serves 20 million daily queries at millisecond latency for 2.3 million daily users, with seasonal resets, tie-breaking, team boards and historical retention, integrated with two worldwide live multiplayer games.
For a US gaming company we broke a Groovy monolith into Go microservices over six months, delivering a transaction engine on EKS with ArgoCD-based GitOps that processes millions of requests per day. For Novomatic we designed a stateless Complex Event Processing engine that analyses 100K+ betting requests per minute and returns player incentives in near real time. Event pipelines like these are core data engineering work.
Game backend cost optimization and reliability
Cost work in gaming is mostly architecture work. We remapped large Terraform stacks into Kubernetes by standing up a parallel environment, load testing it, and shifting traffic with Route53 weighted policies, so the migration caused zero service interruptions and unified runtime stacks cut cloud costs 15 percent.
We also standardise telemetry for gaming clients, including an ongoing engagement with a large global gaming company, because you cannot control cost or reliability you cannot see. That practice is described under telemetry and observability.
Talk to an engineer
