Introduction

Bindu is a next-generation, lossless compression system that makes many kinds of data smaller, and lets you work with that data without decompressing it first.

Traditional tools like gzip, zstd, and brotli only shrink files. To read, search, or change anything inside, you must first decompress the whole file. Bindu allows you to compresses your data, and then read, search, and edit that data while it’s still in compressed form.

We call this novel capability computable compression and it means that you can save money on three fronts at once: lower storage costs, lower bandwidth costs, and lower ongoing compute costs. In addition, Bindu typically compresses data in record time compared to traditional tools.

Bindu capabilities

Bindu unlocks the following key capabilities:

• Industry-leading data compression. Bindu typically matches or beats the best classical compressors on text, source code, structured data, and telemetry.
• Search and edit without decompressing. Find a record in a multi-GB archive in milliseconds, without rehydrating the whole archive. Or change a value in place without the need to first decompress the file.
• One tool for any data. The same Bindu binary handles English text, source code, satellite telemetry, scientific datasets, and structured logs.
• Gets better with use. Bindu remembers patterns it has seen, so each new file you compress benefits from the ones before. You can even have Bindu continually improve without giving it any new data, only additional compute cycles to improve its compression approach.

A concrete example

Suppose you need to replace a single string in a compressed archive of aircraft tracking data in the “Automatic Dependent Surveillance-Broadcast (ADS-B) format.

With a conventional tool like xz, the process is decompress → edit → recompress. In our test environment, that took 1.33 seconds.

With Bindu, you edit the compressed file directly, skipping decompress, and recompress. The same edit took 3 milliseconds, a 443× speedup compared to gzip!

At scale, that speedup translates directly into a smaller compute bill and completing edits faster.