Vector Search easy
Your team proposes an HNSW index for 500 million 1,536-dimension fp32 vectors, but the cluster has a tight RAM budget. What is the main concern?
- A HNSW cannot index more than about 10 million vectors due to a hard limit in the graph format.
- B HNSW requires the vectors to be normalized, which is impossible to do at this scale of data.
- C HNSW returns only exact results, so query latency grows linearly with the number of vectors.
- D HNSW keeps vectors plus graph links in memory, which here means multiple terabytes of RAM.
Answer: D. HNSW keeps vectors plus graph links in memory, which here means multiple terabytes of RAM.
At about 6 KB per vector, the raw data alone is about 3 TB. HNSW graph overhead often adds another 1.5x to 2x, so the index needs multiple terabytes of RAM. HNSW gives high recall at low latency but pays for it in memory; under a tight budget, IVF with product quantization or a disk-based index is the better fit. HNSW is approximate, not exact, so option C misreads how it works.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- You are building an IVF index on 20 million vectors. A teammate asks why the build needs a 'training' step before any vectors can be added. What is the reason?
- An HNSW index serves a marketplace where about 30% of listings are updated or removed each week. Each update is applied as a delete plus an insert, and deletes mark tombstones. Over a month, recall@10 drifts from 97% to 89% and p99 latency rises at the same ef_search. What is happening, and what should you do?
- Users delete documents in your primary database, but the deleted documents keep appearing in vector search results for hours. The vector index is rebuilt nightly. What is the best fix?
- Your embedding model outputs unnormalized vectors whose magnitudes vary widely. A teammate wants to switch the index from cosine similarity to dot product because it is cheaper to compute. What is the consequence?
- You are setting an SLO for a new ANN-backed recommendation search over 50 million vectors. A product manager asks for 100% recall@10 with single-digit millisecond latency. What is a realistic response?
- A B2B SaaS app serves 500 customers, and every query searches only one customer's documents. One large customer runs heavy batch queries that slow down search for everyone. Which sharding approach fits best?