Interview questions in Databases & Architecture, page 5
Connection pools and the problem they solve
A pool reuses pre-opened connections so requests skip the expensive connect handshake; without one, every request pays setup latency and may overwhelm the database.
Pooled connection lifecycle and close() semantics
Borrow from pool, use, then close() returns it to the pool rather than tearing down the socket.
Eager vs lazy loading in an ORM
Eager fetches related data up front (joins/extra query); lazy defers until accessed. Lazy in a loop causes the N+1 query problem.
Transaction isolation levels and their tradeoffs
Isolation levels control which concurrency anomalies (dirty/non-repeatable reads, phantoms) are allowed; higher levels mean stronger consistency but more blocking and less concurrency.
Unit of Work / Session pattern in ORMs
The Unit of Work tracks new, dirty, and deleted objects, then flushes them as one batched transaction at commit.
Fixing an ORM's inefficient aggregation query
Drop to raw SQL or a view for the heavy report, or restructure the ORM query and add indexes. Raw SQL is fast but couples to the schema; tuning keeps portability.
Managed RDS vs self-managed DB on EC2
Managed RDS offloads patching, backups, failover, and replication, freeing the team to build product; self-managed EC2 means you own all that toil.
Read replica vs Multi-AZ in RDS
Multi-AZ is synchronous standby for failover, read replicas are async copies for read throughput, and the two solve different problems.
Replica lag and read-your-writes consistency
Stale reads come from async replica lag, the guarantee a user expects is read-your-writes, and you route that user's reads to the primary after a write.
Aurora Serverless v2 for spiky traffic
Serverless v2 autoscales capacity in fine-grained ACU steps near-instantly, you pay per-ACU-second, and provisioned is fixed cost regardless of load.
Near-zero-downtime database migration to cloud
Assess and provision, do a full load then continuous CDC replication with a tool like DMS, validate, then cut over with a rollback plan.
Cache-aside pattern with Redis and RDS
App checks cache, on miss reads DB and populates, writes invalidate the key, and consistency is eventual.
How Spanner achieves global external consistency
TrueTime gives bounded-uncertainty clocks via GPS and atomic sources, Spanner commit-waits out that uncertainty, and Paxos replicates each shard.
Aurora vs Spanner architecture contrast
Aurora is single-writer with a shared distributed log-based storage and quorum, scaling reads; Spanner shards data with Paxos and TrueTime for global writes.
Multi-region active-passive DR with Aurora
Async global replication to a passive region, promote and repoint traffic on failover, and fence the old primary to prevent split-brain.
Data warehouse vs data lake
Warehouses store structured, schema-on-write data for BI; lakes store raw multi-format data with schema-on-read for exploration and ML.
Schema-on-read in data lakes
Structure is applied at query time not ingest, enabling flexible raw storage and ML, but costing query-time validation and risking data swamps.
Clickstream architecture for real-time and batch
Ingest events into a log like Kafka, fan out to a real-time path for dashboards and a batch path to a lake for ad-hoc analysis.
Exactly-once semantics in stream processing
Exactly-once means each event affects state once despite retries, it is hard because of failures between processing and committing, and you achieve it via idempotency or atomic…
The small files problem in data lakes
Many tiny files create per-file overhead and metadata pressure, hurting scans; fix via compaction, batching writes, and tuning partitioning.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles