Game Networking Weekly #3

, part 3 of 3 in Game Networking Weekly

Peeking into VALORANT's Netcode Wayback

Explains techniques used to mitigate normal netcode peeker’s advantage to promote their tactical shooter’s holder’s advantage design choice.

Rollback Networking in INVERSUS Wayback

A detailed account of adding 4 player peer-to-peer mesh rollback networking to an indie game late in development. Demonstrates real attention to detail around input fairness, visual/audio cues, and desyncs.

Dev Blog - Netcode 2.0 & Next Beta Wayback

An update to their soccer game moves hit decisions to the server. A hit here is defined as any action that makes contact with another player or the ball. This is a netcode redesign after complaints from the playerbase about desyncs and rollbacks. Instead of client hit prediction with server validation, the server now determines the hit and informs the client. They use a kind of rollback netcode, although it’s not described in detail here, and decided to add 3 frames (~50 ms) of local input lag to players’ actions to hide latency and hopefully allow the server hit confirmation to arrive in time to avoid visible corrections.

Leveling Up Networking For A Multi-game Future Wayback

Explains how Riot adapted its backbone to support multiple games with different requirements rather than just Leage of Legends. In each region, multiple points of presence (PoPs) advertise the same public IP address, and anycast delivers client packets to one of them. NAT lets game servers use private addresses and routes return traffic through the same PoP. This moves traffic off the volatile public internet and onto Riot’s more controlled private backbone earlier in its journey. The PoPs also spread and absorb DDoS attack traffic and sharing address reduces the number of public IPv4 address Riot must pay for.