179 questions
No questions match those filters.
← All questions
Inference & ServingMedium· Scaling· Sharding· Replication
Your vector index no longer fits in RAM on one machine....
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansReplication is ineffective for capacity issues because it copies the full index to every node, leaving the per-node memory footprint unchanged. Sharding is the only mechanism that addresses capacity constraints. Calculate the required shard count based on the corpus size (N * d * 4 bytes) plus graph overhead, divided by the available node RAM. Acknowledge that sharding introduces fan-out and impacts tail latency, which must be managed.