Database Comparison Matrix

Interactive strategic framework for selecting state stores based on CAP, access patterns, and operational requirements.

Unlock PDF Cheat Sheet

Workload Driven

Filter by OLTP/OLAP roles to match your access patterns.

Strategic Pros/Cons

Deep-dive into trade-offs for Senior and Staff-level reasoning.

Multi-Select Comparison

Select databases to compare side-by-side.

PostgreSQL

OLTP (can do some OLAP)

Scale

Single node: ~10k–100k simple ops/sec; sharded: 100k–1M+ cluster-wide

Consistency / CAP

Strong ACID, single-node CA; CP-ish if you add sync replication

Core transactional systemspayments

MySQL

OLTP

Scale

Similar to Postgres: tens of thousands TPS per node; sharded: 100k+

Consistency / CAP

Strong ACID per node; CA on single node; CP-ish with strict replication

Web backendspayments

Oracle Database

OLTP (plus built-in OLAP features)

Scale

Engineered to support very high TPS in clusters; think 100k+ tx/sec

Consistency / CAP

Strong ACID; CP within cluster; focus on consistency over availability

Core bankingreal-time payments

IBM Db2 (esp. z/OS)

OLTP

Scale

Benchmarks in high thousands of “business tx/sec”; large banks run much higher with batching

Consistency / CAP

Strong ACID; mainframe-oriented; CP in practice

Core bankingdeposits/loans

CockroachDB

OLTP (distributed SQL)

Scale

Hundreds of thousands TPS with near‑linear scale as nodes added

Consistency / CAP

Strongly consistent distributed SQL; CP by design

Fintech apps needing global consistency and regional data-residency

TiDB

OLTP + some OLAP (HTAP)

Scale

Cluster can handle hundreds of thousands to millions ops/sec across nodes

Consistency / CAP

MySQL-compatible, strongly consistent; CP-style

Real-time analytics on transactional datatrading & risk workloads

Azure Synapse

OLAP

Scale

Optimized for large scans; many complex queries/sec rather than per-row TPS

Consistency / CAP

Analytical engine; consistency at batch/warehouse level

ReportingBI

kdb+ / KX

OLAP (time-series, HFT)

Scale

Millions of events/sec ingest; microsecond-level queries over billions of rows

Consistency / CAP

Strong consistency within process; more like specialized column store than general DB

HFT tick dataorder books

TimescaleDB

OLTP/OLAP hybrid (time-series)

Scale

Hundreds of thousands rows/sec writes; sub-second queries on large series

Consistency / CAP

ACID via Postgres; CA on single node; CP-ish with sync replication

Time-series (market datametrics) with Postgres SQL API

ClickHouse

OLAP (columnar)

Scale

Millions of rows/sec ingest; sub-second queries over billions of rows

Consistency / CAP

Column-oriented analytics; eventual consistency at ingestion; not ACID-focused

Trade analyticsrisk

Cassandra

OLTP (wide-column)

Scale

Millions of writes/sec cluster-wide; very high write throughput

Consistency / CAP

Tunable consistency; AP by default (availability + partition tolerance)

Event logscustomer 360

ScyllaDB

OLTP (wide-column)

Scale

Millions of ops/sec with low p99 latency on modest clusters

Consistency / CAP

Tunable consistency; AP with strong options per operation

Low-latency customer 360fraud detection

MongoDB

OLTP (document)

Scale

Single node: tens of thousands ops/sec; scaled cluster: 100k+

Consistency / CAP

Strong consistency if using majority write concern; CP-ish

Semi-structured customer datacontent

Redis

OLTP-ish (KV)

Scale

Single node: hundreds of thousands to >1M ops/sec; clusters scale out

Consistency / CAP

In-memory store; single-threaded per shard; strong consistency per key in one node

Cachingsessions

Aerospike

OLTP (KV / multi-model)

Scale

Designed for millions of TPS with sub-ms latency

Consistency / CAP

Strong consistency options plus high availability; tunable

Fraud detectionad bidding

Apache Kafka

Log store / streaming, not a DB

Scale

Millions of messages/sec across cluster; very high write throughput

Consistency / CAP

Durable ordered log; not a general queryable store; consistency via offsets

Event pipelinesCDC

Selection Strategy

OLTP vs OLAP

Latency-sensitive transactions vs large scans.

CAP Framing

Choose Consistency (CP) vs Availability (AP).

Persistence

Source of truth vs transient cache.

Access Patterns

Primary Key lookups vs Time-series aggregation.

Interview Tip

"For balances, pick CP; for clickstream, pick AP. Always justify by read/write ratio."