Question 1

How would you design a multi-tenant schema in MongoDB, and what trade-offs exist between separate databases vs shared collections?

Accepted Answer

Use separate databases/collections per tenant or a tenantId field with compound indexes and shard keys; isolate using zones and per-tenant quotas. This question assesses senior proficiency in MongoDB design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 2

Explain storage compaction trade-offs and how you’d schedule them.

Accepted Answer

Compactions reclaim space but add I/O; schedule off-peak, cap concurrency, and separate hot from cold buckets to avoid interference. This question assesses senior proficiency in InfluxDB design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 3

How should you handle a disk failure on a Cassandra node to prevent data loss and downtime?

Accepted Answer

Replace disk/node, bootstrap or replace‑address to stream ranges, run repairs, validate token assignments. This question assesses senior proficiency in Cassandra design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 4

How do you handle unbounded growth of large arrays without hitting document limits or performance issues?

Accepted Answer

Avoid unbounded arrays; use bucketing, sub-collections, or $push with $slice/$sort; index array fields cautiously. This question assesses senior proficiency in MongoDB design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 5

How would you design auditing and compliance controls for sensitive MongoDB workloads?

Accepted Answer

Enable auditing (Enterprise), restrict privileged roles, use field-level encryption/CSFLE where needed, and secure backups/logs. This question assesses senior proficiency in MongoDB design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 6

Design a Redis Cluster keyspace strategy that avoids hash-slot hotspots and enables linear scaling.

Accepted Answer

Use key tags to control slotting (e; , user:{id}:*), randomize shard keys for hot items, and avoid monolithic keys; spread high-traffic prefixes across slots and plan rebalancing via cluster resharding. This question assesses senior proficiency in Redis design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 7

Describe how eviction policy choice affects tail latency for mixed workloads.

Accepted Answer

Allkeys-LFU/LRU can keep hit rate high but cause variance under scans; noeviction minimizes tail latency but rejects writes; choose per SLO and workload. This question assesses senior proficiency in Redis design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 8

What’s your approach to schema evolution when field types change mid-stream?

Accepted Answer

Write new type under a new field key, backfill conversions, migrate queries, then deprecate the old key; avoid in-place type flips. This question assesses senior proficiency in InfluxDB design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 9

How do you analyze an EXPLAIN plan in BigQuery to identify and remove a bottleneck?

Accepted Answer

Find large shuffles/repartitions, push down filters, reduce join keys, and pre-aggregate to cut shuffle volume. This question assesses senior proficiency in BigQuery design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 10

When should you choose on-demand vs flat-rate (Reservations) slots, and what are the trade-offs?

Accepted Answer

On-demand = flexibility/simplicity; flat-rate = predictable performance/cost but requires capacity planning. This question assesses senior proficiency in BigQuery design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 11

How do you design efficient time-series rollups (e.g., hourly/daily aggregates) in Cassandra?

Accepted Answer

Store raw and aggregated tables; use TWCS, windowed partitions, and pre‑compute summaries for common queries. This question assesses senior proficiency in Cassandra design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 12

How do you design for high write throughput in InnoDB/MySQL?

Accepted Answer

Batch writes, use bulk operations, ensure fast storage, tune redo, avoid hot PKs, consider partitioning/sharding. This question assesses senior proficiency in MySQL design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 13

How do you design partitioning for sliding-window tables?

Accepted Answer

Partition by RANGE (date), align local indexes, use EXCHANGE/REORGANIZE PARTITION for fast roll‑in/out. This question assesses senior proficiency in MySQL design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 14

How would you optimize a high-cost query in Snowflake?

Accepted Answer

Use query profile to examine scan, join, and spill; optimize joins, filter early, and cluster if needed. This question assesses senior proficiency in Snowflake design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 15

When should you use REINDEX CONCURRENTLY, and why?

Accepted Answer

Rebuilds an index without blocking writes to the table; used to fix corruption/bloat with minimal downtime. This question assesses senior proficiency in PostgreSQL design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 16

How do partition-wise joins and aggregates work, and when do they help?

Accepted Answer

Planner can push operations to partitions, improving performance if constraints and stats are suitable. This question assesses senior proficiency in PostgreSQL design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 17

How do you scale high-concurrency connections effectively?

Accepted Answer

Use PgBouncer in transaction pooling mode, tune max_connections/CPU affinity, and avoid per-connection overhead. This question assesses senior proficiency in PostgreSQL design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 18

Outline a zero-downtime RDB/AOF backup approach on a busy node.

Accepted Answer

Use filesystem snapshots (LVM/ZFS) of RDB/AOF or copy-on-write friendly storage after BGSAVE; verify checksums and rotate atomically. This question assesses senior proficiency in Redis design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.

Question 19

How would you implement GDPR “right to be forgotten” in Snowflake?

Accepted Answer

Track and delete user data by ID; use tagging to find affected rows/files. This question assesses senior proficiency in Snowflake design and optimization. Strong candidates explain internal mechanisms, provide real-world examples, and analyze performance trade-offs. Candidates who only recite definitions without practical depth signal limited production experience. Follow up with scenario-based questions to gauge true technical depth.