Section 11: The Live-Ops Backlog - Follow-up Feature Evolution
Executive Summary: The interviewer reads six requests off the live-ops backlog: country boards, clans, divisions, rank-change notifications, season history and a rolling 7-day board. Each gets three beats: what changes, what stays the same, and what it costs. Most become one more script or log consumer. Two are traps. A clan total is a sum, which loses the max's safety, and a rolling window has to undo a max.
Interviewer: "Launch went well. Live-ops has a backlog. What does each item cost, and what does it break?"
Under all six requests sits one question: is the new thing still a max? If it is, retries, replays and rebuilds stay free behind Section 7's board applier. If it isn't, say so first.
The Fan-Out Map
Only country boards touch the synchronous script. Everything else hangs off data the design already keeps.
The Best-Score Writer gains one output. Its conditional write already sees old and new bests, so it publishes a best-improved event (playerId, boardId, old, new, runId). At peak that's the ledger's ≈ 4,350/s, of which 3,480/s is weekly. Once clans and pushes hang off this topic, Writer lag becomes user-visible, so it gets a lag SLO of ≤ 1 minute with an alert, the bound Section 7's rebuild RTO already relies on.
Sign in to continue reading
The rest of this lesson is available with a free account. Signing in with Google or Microsoft is free.
Sign in to read the full lesson