Skip to main content
All comparisons/ScramDB vs Oracle Database

ScramDB vs Oracle Database

Oracle can run columnar analytics, through Database In-Memory: an Enterprise Edition option, licensed separately, and required for every processor on the server. Enterprise itself is licensed per processor with a core factor, so a modern server counts for more than you expect. The free edition stops at 2 CPU cores, 2 GB of memory and 12 GB of data. ScramDB is columnar for everyone, uses every core you own, and its free tier holds 64 GiB.

Columnar analytics
Included
Oracle’s in-memory columnar store is a separately licensed Enterprise Edition option.
Free tier
64 GiB
Oracle’s free edition stops at 2 CPU cores, 2 GB of RAM and 12 GB of user data.
Cores
All of them
No core factor, no per-processor licence, no counting sockets.
Analytics
31.6s
A full analytical workload over 100 million rows, while transactions run on the same copy.
At a glance

Side by side

ScramDBOracle Database
Columnar analyticsIn the engine, for every userDatabase In-Memory: an extra-cost Enterprise Edition option, licensed for all processors on the server
LicensingCommunity free up to 64 GiB stored. Enterprise for clustering and beyondPer processor with a core factor, or named user plus with minimums, per edition
Free edition ceiling64 GiB of stored data, every core you have2 CPU cores, 2 GB of RAM, 12 GB of user data
Row-level securityPolicies in the engine, in every editionVirtual Private Database, bundled with Enterprise Edition only
Wire protocolPostgreSQL wire protocol, every standard driver and BI toolOracle Net, with its own client stack
TransactionsFull ACID up to serializable isolation, readers never block writersFull ACID with multiversion read consistency
BranchingFork the database instantly, or open it as it was at any past timestampFlashback, RMAN and pluggable database cloning, per edition
AI and agentsA semantic layer inside the engine that agents connect to directly, bound by the caller’s own permissionsNative vector type and AI Vector Search, included in Standard and Enterprise

You already know the licensing story

Oracle prices by processor, with a core factor that turns physical cores into billable processors, and the analytical engine is an option on top of the edition that allows it. Every capacity decision becomes a procurement decision.

ScramDB is a binary. The columnar engine, the parallel execution, the row-level security and the agent interface are the product, not a tier of it.

Everything the application needs, on the same copy

PL/SQL procedures become PL/pgSQL. Triggers stay triggers. Roles, grants, foreign keys and row-level security all carry across, on the PostgreSQL wire protocol that every modern driver already speaks. And the analytical query runs against the live rows without an in-memory option to license first.

When Oracle is the right answer

Decades-deep PL/SQL estates, RAC, Exadata, and regulatory environments where Oracle support contracts are part of the architecture.

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.