DRAG DROP (Drag and Drop is not supported)A recommendation system stores 3 million embeddings in Azure Database for PostgreSQL.During peak hours, P95 similarity-query latency increases, and the cache hit ratio drops significantly.You suspect that the vector index working set no longer fits in memory, resulting in more disk reads.You need to scale server resources and validate the outcome.Which four actions should you perform in sequence? To answer, move the appropriate actions from the list of actions to the answer area and arrange them in the correct order.Select and Place:
Answer(s): A
Step 1: Before peak hours, run a fixed workload to capture baseline P95 latency and capture EXPLAIN ANALYZE for the vector queryEstablishing a baseline is crucial for performance tuning. Running EXPLAIN ANALYZE provides deep insight into whether the query is actually using the vector index (like HNSW or IVFFlat) or falling back to a slow sequential scan.Step 2: Scale to a tier with more memory per vCore Since the problem states that the vector index working set no longer fits in memory, scaling to a tier with higher memory (such as Memory-Optimized tiers) directly solves the root cause by allowing the RAM to fully cache the vector index, eliminating slow disk reads.Step 3: Re-run the same benchmark and compare P95 and P99 latency To validate the outcome of your scaling operations, you must run the identical benchmark workload and compare high-percentile latencies (P95, P99) against your baseline to prove the bottleneck was resolved.Step 4: Review memory pressure and cache hit ratio during peak hours and compare them to the baseline This step confirms the success of the resource scaling. A successful remediation will show a significantly increased cache hit ratio and reduced memory pressure/disk I/O compared to the baseline metrics.
https://www.pgedge.com/blog/ai-features-in-pgadmin-ai-insights-for-explain-plans
You are implementing the expiration requirement for AI-generated summaries in the Redis cache.You need to purge the AI summaries five minutes after creation, regardless of how often they are accessed.What should you do?
Answer(s): B
TTL (Time-To-Live) with a fixed duration of five minutes is the exact Redis feature you need to purge AI-generated summaries regardless of access frequency.Redis TTL and Expiration Features Core Expiration Feature EXPIRE command: Sets a timeout in seconds on a key. Use EXPIRE key 300 to remove the summary after 300 seconds (five minutes).PX command: Sets a timeout in milliseconds. Use PX key 300000 for precise millisecond control.EXPIREAT / PEXPIREAT: Sets an absolute Unix timestamp for expiration instead of a relative duration.
https://dev.to/truongpx396/30-common-redis-interview-questions-1ai
You optimize an AI inference API that uses Redis caching.You must reduce the risk of serving outdated data while minimizing cache management overhead.You need to implement the caching strategy that satisfies the requirements.What should you do?
Trigger invalidation when source data changes is the best action.Data Freshness: This is a proactive cache-aside or write-through approach. By explicitly deleting or updating the specific Redis key when the underlying database or model data changes, you completely eliminate the risk of serving stale data.Low Overhead: It keeps cache management overhead minimal because invalidation logic is execution-dependent (event-driven). Instead of constantly updating expirations or relying on heavy computing logic, you only trigger an action when a write/change event occurs.Incorrect: [not B] Set eviction policy to allkeys-lru: This policy is a memory management feature, not a data-freshness strategy. It only deletes the Least Recently Used (LRU) keys when your Redis instance runs out of RAM. If memory usage is stable, outdated data will sit in the cache forever, posing a major risk of serving stale inferences.[not D] Reset expiration on each read (Sliding TTL):This actively increases the risk of serving outdated data. If a specific AI response is highly popular and read continuously, its expiration window will extend indefinitely. The data remains stuck in the cache even if the source data undergoes critical updates.
https://oneuptime.com/blog/post/2026-01-30-llm-caching-strategies/view
You are implementing semantic retrieval for a chatbot.Embeddings are already stored in Redis. However, vector similarity queries do not return matches.You need to resolve the vector similarity search issue.What should you do?
Answer(s): C
Creating a FLAT vector index on your embedding field will resolve the issue and allow your Redis vector similarity queries to return matches.In Redis (using the Redis Search and Query features), vector fields cannot be queried using Vector Similarity Search (VSS) syntax until a dedicated vector index is explicitly built over them.No Automatic Indexing: Redis does not automatically index JSON or Hash fields containing raw binary or string embeddings.Query Failure: Without an index, VSS queries (using the KNN operator) will fail with syntax errors or return zero results because the query engine cannot parse the unindexed field.
https://www.louisbouchard.ai/indexing-methods/
HOTSPOT (Drag and Drop is not supported)You are building a semantic search feature for a chatbot. You store document embeddings in Redis.You review the following Python code that connects to Redis and stores an embedding value:For each of the following statements, select Yes if the statement is true. Otherwise, select No.Note: Each correct selection is worth one point.Hot Area:
Box 1: Yes The line r.hset("doc:1", mapping={"embedding": embedding}) will store the embedding as a field within a Redis Hash.Box 2: NoBox 3: Yes
https://redis.io/docs/latest/commands/expire/
Post your Comments and Discuss Microsoft AI-200 exam prep with other Community members:
Our website is free, but we have to fight against AI bots and content theft. We're sorry for the inconvenience caused by these security measures. You can access the rest of the AI-200 content, but please register or login to continue.