Question 1

What are sharding and partitioning?

Accepted Answer

Both split a large dataset into pieces. **Partitioning** divides a table **within one database**; **sharding** distributes data **across separate database instances**. The distinction matters because one is an optimisation and the other changes your architecture fundamentally.

**Partitioning.** A single logical table is stored as several physical pieces, by range, by list, or by hash of a column. The database handles it, queries are unchanged, and the benefits are real: **partition pruning**, where the planner skips partitions that cannot contain matching rows; cheap bulk deletion by dropping a whole partition; and smaller indexes per partition.

**Its limit:** you are still bound by one machine's capacity.

**Sharding.** Data is divided across independent databases, each holding a subset, routed by a **shard key**. This buys horizontal scale — and costs a great deal.

**What sharding actually costs:**

**Cross-shard queries become expensive or impossible.** Anything that must touch all shards is now a scatter-gather, and joins across shards generally are not supported.

**Transactions across shards** require distributed transaction protocols, which are slow and complex — so most systems avoid them by designing so transactions stay within a shard.

**The shard key becomes an irreversible decision.** Choosing badly produces **hot shards** where one holds disproportionate traffic, and changing it later requires migrating everything.

**Rebalancing is genuinely hard**, which is why consistent hashing and virtual shards exist — to move a fraction of data rather than remapping all of it.

**Operational burden multiplies**: backups, schema migrations, monitoring and failover, per shard.

**Global uniqueness and ordering** become non-trivial, which is why distributed ID schemes exist.

**What to do before sharding**, because it is very frequently unnecessary: add indexes, fix the queries, add caching, use read replicas, archive cold data, partition within one database, and buy a larger machine — **vertical scaling reaches much further than people assume.**

**Shard last**, when you have measured that nothing else suffices.

Question 2

What techniques can I use to tune the performance of MongoDB?

Accepted Answer

MongoDB performance tuning techniques include indexing, sharding, schema design, connection pooling, query optimization, and hardware optimization. Each method enhances database efficiency and responsiveness.

Question 3

What is a hash table, and why is lookup constant time?

Accepted Answer

A structure that turns a key into an array index by computing a number from it, so finding an item takes roughly the same time whether the table holds ten entries or ten million. The mechanism. A

Question 4

When does a database index hurt performance?

Accepted Answer

Indexes speed up reads by maintaining an ordered structure alongside the table, and that structure has to be kept correct on every write. Every insert, update and delete must update every affected

Question 5

What is the difference between scaling up and scaling out?

Accepted Answer

Scaling up means a bigger machine; scaling out means more machines. They fail at different points and impose completely different demands on your software. Scaling up — vertical. More cores, more

Question 6

What is the difference between concurrency and parallelism?

Accepted Answer

Concurrency is about structure; parallelism is about execution. A concurrent program is composed of tasks that can make progress independently; a parallel program actually runs them at the same

Question 7

What does ACID mean in databases?

Accepted Answer

ACID names four guarantees a database provides around transactions. Atomicity means all or nothing. A transaction that updates several rows either applies completely or leaves no trace. A crash

Related Questions