whichdb

844 profiles · independent · open data

Find the right database for what you're building.

Describe your project in plain English. WhichDB turns it into requirements, scores every database against them and shows its reasoning. No sign-up, no sponsored rankings.

describe workload, scale, cloud, license…
databases profiled
800+
data models
19
use cases
33
open source
57%

rankings

Most popular databases

WhichDB Score: a transparent 0–100 index of current developer adoption and mindshare.

Full leaderboard→
  1. 1.PostgreSQLby PostgreSQL Global Development Group93.2Relational
  2. 2.Redisby Redis Ltd.84.3Key-Value
  3. 3.MySQLby Oracle Corporation82.3Relational
  4. 4.MongoDBby MongoDB, Inc.80.5Document
  5. 5.SQLiteby D. Richard Hipp / Hwaci (SQLite Consortium)75.5Relational
  6. 6.Elasticsearchby Elastic N.V.75.1Search Engine
  7. 7.Snowflakeby Snowflake Inc.74.4Columnar / OLAP
  8. 8.Microsoft SQL Serverby Microsoft71.0Relational
  9. 9.Prometheusby Prometheus Authors (CNCF)70.8Time-Series
  10. 10.ClickHouseby ClickHouse, Inc.69.8Columnar / OLAP

rising

Mindshare above their size

Talked about far more than their install base suggests. Worth a look before they're mainstream.

SigNoz

by SigNoz Inc.

OpenTelemetry-native APM and observability platform storing logs, metrics and traces in ClickHouse.

WhichDB Score17.0HN mentions / yr79

TigerBeetle

by TigerBeetle, Inc.

Purpose-built financial transactions database: double-entry accounts and transfers at very high throughput.

WhichDB Score19.2HN mentions / yr257

Dolt

by DoltHub

MySQL-compatible SQL database you can branch, diff, merge and clone like a Git repository.

WhichDB Score18.2HN mentions / yr81

use cases

Start from the job

Purpose-built options first, then the generalists that handle it well.

RAG & AI memoryRetrieval for LLM apps: embeddings, hybrid search, filtering and agent memory.68 →Transactional apps (OLTP)Many small reads and writes with transactions: the backbone of most applications.212 →Real-time analyticsSub-second queries over streaming, constantly-arriving data for user-facing dashboards.92 →CachingKeep hot data in memory in front of a slower system of record.48 →Time-series dataMetrics, sensor readings and events indexed by time.62 →Mobile & offlineOn-device storage that keeps working offline and syncs later.33 →Full-text searchKeyword search with ranking, stemming, facets and typo tolerance.60 →Data warehouseCentral store for analytics across the business, usually SQL and columnar.48 →Graph traversalMulti-hop relationship queries in milliseconds (friends-of-friends, paths).31 →Edge & serverlessData close to users, accessible from edge functions with low latency.42 →Stream processingContinuous queries and incrementally maintained views over event streams.26 →Embedded / in-processRuns inside your application process with no server to operate.101 →

heads up

Sunset & deprecated

how it works

Reasoning you can check

01 Describe it

  • Plain English: “multi-tenant SaaS on AWS, Postgres compatible, global”
  • Or pick requirements by hand

02 See what was understood

  • Your words become explicit requirements
  • Mark each one nice-to-have or must-have
  • Add or remove any

03 Get explained matches

  • Every database scored on every requirement
  • ✓, partial or ✗ for each, never a black box
  • Compare the finalists side by side
Try smart match →

faq

Questions

How is the WhichDB Score calculated?

It is a 0–100 popularity index built from public signals, weighted toward the last 12 months: client-library downloads, Hacker News discussion, job posts, GitHub star growth, Wikipedia pageviews, Docker pulls and, for databases released by 2021, Stack Overflow questions. Each signal is log-scaled and weighted; signals that cannot apply (no public repo for a closed product) are left out rather than counted as zero. The methodology page has the exact weights.

Why doesn't Stack Overflow decide the ranking?

Stack Overflow's question volume has fallen about 99% since its peak as developers moved to AI assistants, so recent activity there says little about adoption. Its all-time question count still reflects installed base, so it keeps a modest weight for databases released by 2021.

How does smart match pick databases?

Your description is parsed into explicit requirements (workload, data model, scale, deployment, cloud, license, compatibility). Every database is scored per requirement with partial credit, must-haves gate the result, and popularity only breaks ties. You see the reasoning for each match.

Is WhichDB affiliated with any vendor?

No. Profiles are researched from official docs, repositories, release notes and papers, and every profile lists its sources. Nobody pays for placement.

Why list dead and research databases?

Because people still search for them, run them, or need to migrate off them. Every discontinued or research system is clearly marked and links to its modern successors.

Something is wrong or missing. How do I fix it?

Each profile has a “Suggest an edit” link to its JSON file on GitHub. Profiles are plain data, so corrections are quick to review.