rumi
01

The problem

The same flexibility that makes GeoTIFF universal makes it expensive at large scale.

Flexibility is the enemy of scale

Asterisk

Every tile needs an offset and a size

Asterisk

Each codec is a separate integration

Asterisk
02

The approach

Fixed container. Direct frame access.

One container profile, two frame units

Asterisk

The index is a prefix sum

Asterisk

Gigabytes of index become megabytes

Asterisk

Compression is a graph, not a monolith

Asterisk
03

Benchmark

Measured on a real Sentinel-2 band — almost the best ratio, a fraction of the cost.

Benchmark I

Asterisk

Benchmark II

Asterisk
reproduce every number on Google Colab
04

The data model

rumi is one stone in a larger structure.

rumi is part of the khipu data model

Asterisk