The Ride-Hailing Hotspot on New Year's Eve
The Ride-Hailing Hotspot on New Year's Eve
Your ride-sharing app tracks live GPS coordinates for 400,000 active drivers worldwide, receiving location pings every 4 seconds (100,000 GPS writes/second). Passenger apps query "find the 10 nearest drivers within 2 km" at 35,000 read queries/second. The existing system partitions drivers into a Redis cluster using a fixed 6-character Geohash (~1.2 km x 0.6 km cell). On New Year's Eve at midnight in Manhattan (Times Square), driver and passenger density surges astronomically: 12,000 drivers and 45,000 passengers converge inside a single Geohash cell (dr5ru6). The single Redis node responsible for the hash key of cell dr5ru6 pegs at 100% CPU on its single-threaded event loop. Nearby search queries for Manhattan timeout, while neighboring rural cells with zero drivers waste memory and CPU. Furthermore, riders situated near the edge of a cell fail to see drivers parked 50 meters away across the cell boundary. You are asked to redesign the geospatial partitioning scheme to adapt to extreme density variations and eliminate edge-boundary query blind spots.
The Ride-Hailing Hotspot on New Year's Eve
Your ride-sharing app tracks live GPS coordinates for 400,000 active drivers worldwide, receiving location pings every 4 seconds (100,000 GPS writes/second). Passenger apps query "find the 10 nearest drivers within 2 km" at 35,000 read queries/second. The existing system partitions drivers into a Redis cluster using a fixed 6-character Geohash (~1.2 km x 0.6 km cell). On New Year's Eve at midnight in Manhattan (Times Square), driver and passenger density surges astronomically: 12,000 drivers and 45,000 passengers converge inside a single Geohash cell (dr5ru6). The single Redis node responsible for the hash key of cell dr5ru6 pegs at 100% CPU on its single-threaded event loop. Nearby search queries for Manhattan timeout, while neighboring rural cells with zero drivers waste memory and CPU. Furthermore, riders situated near the edge of a cell fail to see drivers parked 50 meters away across the cell boundary. You are asked to redesign the geospatial partitioning scheme to adapt to extreme density variations and eliminate edge-boundary query blind spots.
Provide 1–2 precise sentences for each architectural dimension. Each box guides you on what staff-level interviewers evaluate.
Define SLA targets, hard consistency constraints, and conditions the system must never violate.
Quantify throughput (QPS/RPS), read:write ratios, and peak burst multipliers.
Step-by-step path: client ingress → API gateway → queues → background workers → persistence.
Database engine, table schema, partition keys (PK/SK), and durability strategy.
What resource hits saturation first under 10x traffic? (CPU, disk IOPS, connection pools, network).
Worker crashes, network partitions, split-brain, poison pill DLQ, retries, and idempotency.
What did you sacrifice in exchange and why? (e.g. eventual consistency vs latency, cost vs redundancy).