Rematch | Sloclap
How Multiplayer Wizards engineered a reversible, zero-downtime migration to Amazon GameLift Servers
“It was just insanely smooth and seamless.”
— Victor Watrelos, Online Programmer, Sloclap
Executive Summary
Sloclap needed to replace the game-server hosting behind Rematch while the game remained live. Multiplayer Wizards designed the migration around one engineering principle: every step had to be observable, measurable, and reversible, so the team could change direction without putting players at risk.
The deadline was tight. Sloclap had four months to find, adapt, test, and move to a new hosting solution. Rather than treat that as a single cutover, Sloclap, Multiplayer Wizards, and AWS kept the old and new environments running in parallel, built using a percentage-based regional routing system, validated capacity and instance choices, expanded monitoring, and prepared rollback paths before production traffic moved.
The migration was completed in under three months. On cutover day, a stable 10% initial production cutover test gave the team enough evidence to keep increasing traffic until 100% of players were on Amazon GameLift Servers that same day, with zero downtime, servers continuing to run at 60 FPS, and hosting costs reduced by 30–40%.
| 100% of players migrated with zero downtime | 3 months to complete migration | 60 FPS server performance maintained | 30-40%reduction in hosting costs |
Game & Studio Context
Rematch is Sloclap’s competitive multiplayer football game, built around fast, real-time play where reliable, low-latency connections are part of the experience. The game was already live, so changing the infrastructure underneath it could not come at the expense of the people already playing.
The trigger arrived in December 2025, when Sloclap learned that the provider hosting Rematch’s game servers was being acquired and would transition to a new company on a tight timeline. That turned a future infrastructure decision into an immediate engineering problem: replace a critical live service, prove the replacement, and move production traffic within four months.
The Challenge
From the player’s point of view, the best possible migration was one they never noticed. Rematch needed to stay online and responsive while traffic moved between hosting environments, with game servers continuing to run at 60 FPS.
From the engineering team’s point of view, the move also needed to improve the operating model. Sloclap’s previous hybrid setup combined bare metal with third-party cloud capacity, but limited access to the machines running the game and made custom tooling, debugging, and operational improvements harder than the team wanted.
That made reversibility the central design requirement. The team needed to run both environments in parallel, control how much traffic reached each one, watch the right signals in real time, and retain a genuine route back if a production wave behaved differently from testing.
Build the Escape Route First
Multiplayer Wizards approached the project as a controlled live-systems change rather than a simple hosting switch. The team had already worked with Sloclap on a technical evaluation of Rematch before launch, while AWS brought together more than 40 people across account, solutions architecture, and service teams to support the migration.
Multiplayer Wizards embedded an engineer with Sloclap early in the engagement and established clear operational ownership before cutover. During the migration, a 24/7 command centre kept Sloclap, Multiplayer Wizards, and AWS connected, followed by around-the-clock support for three months after the move.
Two environments, one controlled route.
The key architectural decision was to keep Sloclap’s previous provider and Amazon GameLift Servers running in parallel. Sloclap built a routing system that could direct a chosen percentage of player allocations, by region, to either environment.
That turned rollback from an emergency idea into part of the normal migration mechanism. Traffic could move forward in controlled increments, but it could also be redirected if the data showed a problem. Multiplayer Wizards supported the design with region planning, operational runbooks and full observability stack, and the criteria needed to decide whether to continue, pause, or reverse a wave.
Prove It, Watch It, Then Move It
Reversibility made the migration safer, but the new environment still had to earn production traffic. Multiplayer Wizards supported capacity planning, region selection, instance-type testing, performance validation, and right-sizing so Sloclap could understand how Amazon GameLift Servers would behave before relying on it at scale.
Observability was treated as part of the migration architecture, not an afterthought. Sloclap’s backend already ran on Amazon ECS and used Amazon GameLift Servers FlexMatch for matchmaking; Amazon CloudWatch metrics were extended into custom dashboards and alarms so the team could see game-server health clearly and react quickly during each production step.
10% first. 100% when the data said go.
Sloclap completed the migration on 21 February 2026. The rollout had been designed to take as long as a week, but the first 10% production wave proved stable. Instead of following the calendar for its own sake, the team followed the evidence, raising the percentage repeatedly until 100% of player allocations were on Amazon GameLift Servers that same day.
The cutover completed with zero player downtime while Rematch servers continued running at 60 FPS. The migration mechanism had been deliberately cautious; the result was fast because the telemetry kept showing that it was safe to proceed, not because the team removed safeguards.
Results & Benefits
The most important outcome was more than a quick finished migration; the team wanted one that players never noticed was happening. Sloclap moved 100% of Rematch players to Amazon GameLift Servers with zero downtime, completed the wider project in under three months, and maintained 60 FPS server performance through the cutover.
The new environment also reduced hosting costs by 30–40% compared with the previous provider and gave Sloclap more direct control over its game-server estate. Engineers can connect to machine consoles, attach debugging tools, run local servers integrated into the Amazon GameLift Servers flow, and build custom tooling that had been difficult to maintain before.
| Area | Before | After |
|---|---|---|
| Player migration | Production traffic on previous hosting | 100% migrated with zero downtime |
| Cutover pace | Staged rollout allowed up to one week | 10% to 100% on the same day |
| Observability | Limited visibility and machine access | Custom dashboards, alarms, and direct access to machines |
| Hosting costs | Previous hosting cost baseline | 30–40% lower hosting costs |
| Migration support | Urgent live-service deadline | Embedded engineer, 24/7 command centre, plus three months’ support |
“With AWS, if something goes wrong, we detect it way faster than before.”
— Victor Watrelos, Online Programmer, Sloclap
Why It Mattered: A Better Operation Model, Not Just a New Host
Replacing the previous provider solved the immediate deadline, but the longer-term value came from changing how Sloclap could operate Rematch. Direct machine access and stronger debugging options removed some of the constraints of the old hosting model, while custom dashboards and alarms improved the team’s visibility into game-server health.
The networking setup also became simpler; Sloclap had previously used Network Next to improve routing for some players with poor connections. After migration, the studio removed it because the AWS network reduced the number of players affected by those connection issues.
In this way, Sloclap did not trade operational control for lower cost, or stability for speed. The migration delivered a 30–40% hosting cost reduction alongside better access, faster detection of problems, more useful tooling, and a cutover that remained invisible to players.
“I think the biggest impact is the fact that it worked so well that there wasn’t really an impact.”
— Victor Watrelos, Online Programmer, Sloclap
What We Learned
The Rematch migration shows that the safest way to move a live multiplayer service quickly is to engineer the ability to stop. Parallel environments, controllable traffic, production telemetry, and clear operational ownership gave the team freedom to accelerate because every step remained measurable and reversible.
| Build rollback before momentum: A migration is safer when the route back exists before the route forward is used. |
| Give traffic a dial, not a switch: Percentage-based routing made production migration incremental and controllable. |
| Let telemetry set the pace: A stable 10% wave – not the calendar – gave the team confidence to accelerate to 100%. |
| Judge the post-cutover operating model: The best migration leaves the team with better access, tooling, observability, and economics. |
Subscribe to our
newsletter
"*" indicates required fields