For true latency reduction, the only answer is more locations
Jul 25, 2023
5-10 data centers don’t actually enable game devs to deliver low latency to players regionally while maintaining ultra-low latency between regions. The only solution is to access a global, public edge computing infrastructure. However, this is very hard to manage, so it requires optimized, automated orchestration to handle it and ensure game server deployments in the optimal locations for all its players.
The “Region” Challenge & Limitations
When looking to provide low latency experience worldwide, the default response is to pick more than 1 location. That’s common sense right there as more locations mean closer to players, thus shorter distance and faster round trip time.
Studio’s technical directors or CTO will typically pick a handful of regions to cover worldwide, 1 or 2 on USA’s East Coast. and the same for its West Coast. They’ll likely add in one in Europe, one in APAC (if at all), and call it a day. The general belief is that this suffices in proving sub-40ms latency worldwide and adding a “fast backbone” will mitigate people far away.
There are two drivers for such a decision:
1. The cost of hosting and managing more than 1 region and;
2. Managing potential matchmaking issues by splitting the world into multiple sub-regions limiting the distance of potential players by filtering their locations
Managing servers costs is necessary for all studio sizes, and this challenge directly impacts the player’s experience by limiting the number of available locations for players to play on their game servers. As unfortunately, the direct consequence is an increase in lag for a game’s player base because, by definition, such a limited number of locations for players to play on.
Global Distribution & Increased Number of Locations
Those two arguments are not saying that more locations would reduce latency, they are saying that such solution is too expensive and may cause problems elsewhere.
Providers like Edgegap remove the cost problem by offering multi-tenancy, just-in-time game hosting in hundreds of regions without committing to a specific set of regions.
This is proven by Edgegap’s case study with a AAA publisher. Despite having a large number of locations (more than what even most studios would be able to afford!), by using traffic from 600,000 transactions and comparing the results with a AAA studio’s current architecture, Edgegap demonstrated an average latency reduction from 116 milliseconds to a drastic 48 milliseconds. On top of that, 78% of the transactions had a latency below 50 milliseconds, compared to only 14% without the Edgegap solution.
The other problem is due to the old architecture of the studio’s matchmaker. A studio can easily remove physical regions in matchmaking, resulting in a single region for matchmaking. Using latency along with LOM allows for good matchmaking, while still allowing people from different region to play together.
The key here is to have a smart orchestrator after the matchmaker to pick the best location out of many, for each match. The result is extra-proximity matches with sub-10ms latency for every player, without impacting matchmaking queuing time.