Skip to main content
All comparisons/ScramDB vs Databend

ScramDB vs Databend

Databend is an elastic warehouse over object storage, which means a Raft-based meta service for consistency, stateless query nodes for compute and an S3-compatible bucket for the data: three components before the first query. It is also downstream of your transactional database rather than being it. ScramDB answered the same analytical workload faster from a single binary, and holds the transactions on that same copy.

Faster
1.3x
ScramDB 31.6s against Databend 42.3s. Both answered every query.
Smaller on disk
1.2x
ScramDB 16.1 GiB against Databend 19.5 GiB. Same data, after load.
At a glance

Side by side

ScramDBDatabend
What you deployOne binary, one processA Raft meta service, query nodes and an object store
Transactional workloadServed by the same engine on the same copyA warehouse, fed from your database
IsolationRead committed, repeatable read and serializableMulti-statement transactions with the isolation level not clearly documented
Referential integrityForeign keys and CHECK constraints, enforcedNot documented as enforced
Latency floorLocal disk, no object store in the pathObject storage, with a metadata service for consistency
Wire protocolPostgreSQL wire protocol, every standard driver and BI toolIts own HTTP handler and MySQL protocol
BranchingFork the database instantly, or open it as it was at any past timestampSnapshot based time travel
AI and agentsA semantic layer inside the engine that agents connect to directly, bound by the caller’s own permissionsNative vector type with an HNSW index, a genuine strength

Object storage is a cost as well as a feature

Separating storage from compute buys elasticity and adds a network hop to every read, plus a metadata service to make an eventually-consistent bucket behave transactionally. For a warehouse that runs batch reports, that is a good trade. For the database behind an application, it is latency you cannot get back.

When Databend is the right answer

Elastic warehouse analytics directly on object storage under Apache 2.0, where compute needs to scale to zero between workloads.

The measurement, as of 5 September 2026. A full analytical workload over 100 million rows of real web analytics data, run on a 16 vCPU AWS instance. Ours is a single ScramDB node in its stock configuration, with the GPU off and nothing tuned for the test. Rival figures are those systems’ own published results for the same instance type. These are the numbers as of that date, not a permanent scoreboard: the engine is tuned continuously and it has got faster through every round of this work, so a later run should read better than this one.

Run it yourself in one line

One static binary, no dependencies. It picks its own config, listens on the postgres port 5432 and runs in the background.

$curl -fsSL https://scramdb.com/install | bash

Quick start · How the engine works · Talk to us

A full analytical workload over 100 million rows of real web analytics data, run on a 16 vCPU AWS instance, measured 5 September 2026. Ours is a single ScramDB node in its stock configuration, with the GPU off and nothing tuned for the test. Rival figures are those systems’ own published results for the same instance type. These are the numbers as of that date, not a permanent scoreboard: the engine is tuned continuously and it has got faster through every round of this work, so a later run should read better than this one.