We believe that backups should be easy, fast and last forever.
Asuran is a new archive format and rust implementation. It aims to be the archiver for the 2020's, and has been written from the ground up to use the insights from cutting edge research and extract every last bit of performance out of modren hardware, while still providing features users have come to rely on in archivers, like encryption, compression, and global deduplication.
Asuran should be sutible for the long term archival of data, should be operating system and hardware independent, secure, flexible, fast, and easily embeddable.
It should strive to make backups a fast and easy process, and to allow the user to preserve as much of their file history as possible in the space they have avaible. After all, what good is a backup that never ran because it would have taken too long, or the backup that got deleted because it was using too much space?
Asuran should be safe for use on untrusted storage, and should not leak any data that could reveal, to any extent, the contents of the repository.
How Does it Work?
Asuran works by splitting your files up into a number of chunks. It splits files up using a selectable Content Defined Chunking algorithim, currently FastCDC by default, so that even if one part of a file changes, it is very likely the other chunks will not..
Those chunks are then (optionally, but on by default) compressed and encrypted, before being comitted to a Content Addressable Storage backend, for later retrievial. Asuran will try its best to only store each chunk it encounters once (and in the most common usecases, it can do this 100% of the time), preventing an archive from storing the same information in the repository more than once if it can avoid it.
The entire archive structure is verified with a Merkele Tree process, so you can be sure that if your restore is successful, then your data is intact and has not been tampered with.
You can store as many archives as you want in a single repository, and write to the same repository with as many computers as you want.
The entire data pipeline is built on top of a modren, async stack, allowing performance that previous contentors in this space could only have dreamed of.
Installing and using
In most cases you will be interacting with the command line asuran clinet (asuran-cli). Either build it from source from the asuran-cli directory in this repository, or install it with:
env RUSTFLAGS="-C target-feature=+aes,+ssse3" cargo install asuran-cli
-C target-cpu=native for even better performance. The target features (aes and sse3) are required to get good performance, and asuran does not currently offically support being built without them.
asuran-cli --help for usage.
asuran-cli is, at heart, a thing wrapper that glues togehter the API of the
asuran library. The
asuran crate provides a high level interface for interacting with repositories, and will always be a sepereate component and enjoy the same level of support as
Basic Overview and Terminology
The asuran format is split into three logical layers
Backend and Repository are used somewhat interchangeably. This is the actual place where data gets stored. It is a content addressable storage backend, where blobs (called
Chunks) are addressed by an HMAC of their plaintext
This is a data structure, stored in a repository, that describes the way chunks are stitched together to form objects/files. If you think of a repository like a time machine backup, archives are your snapshots
The manifest is a special data structure stored outside, but adjacent to, the other parts of the repository. It maintains pointers to all the archive structures and provides the root of verification for all the data in the repository.
Comparison with Borg
Asuran Has better use of multithreading
Asuran has a pipeline that is much better suited to good use of CPU cores, and since archiving speed is typically bound by compression, rather than read/write or encryption/hmac on modren CPUs with an SSD (or just fast spinning rust storage), this leads to a significant speed boost.
Switchable storage backends
Asuran has a framework for describing new storage backends and switching between them at runtime, so you aren't tied to only storing files on a local filesystem or linux machines running ssh
Asuran allows you to chose your content defined chunking algorithm, including the use of FastCDC or a static chunk size slicer, allowing you to decide on your performance/deduplication ratio trade off
Repository format hides chunk length
Asuran does not suffer from chunk length based fingerprinting attacks like borg does, since we use a repository format that hides chunk length.
Comparison with Restic
Asuran is much faster
Asuran is generally faster than borg, and borg is generally faster than Restic, so this one just follows
Asuran supports multiple cipher suites of roughly equivalent security, allowing you to meet organizational requirements to use a specific cipher, or to select the cipher that runs fastest on your hardware. You can also completely disable encryption in the event that you actually are backing up to trusted storage.
Support for compression
Asuran lets you chose between no compression, ZStd, LZMA, or LZ4 compression for your repository, allowing you to choose your time/space trade off
Comparison with Rdedup
Built in directory traversal
Asuran will traverse a directory and store its structure automatically. No need to perform an extra step and decrease deduplication by creating a tarball beforehand
Multiple backend support
See the above note in the borg section
Improvements To All
High level api suitable for embedding
Asuran presents a high level api that is consumed by
asuran-cli, making it easy to embed support for asuran archives into other applications
Multiple input format support
Asuran is agnostic to where the files it is backing up are coming from. It doesn't have to be from a filesystem, asuran could just as easily import individual files from a tarball or directly dump tables from a database.
Asuran is distributed under the terms of the BSD 2 Clause + Patent License.
By contributing to this project, you agree to license your contributions under the terms of the BSD 2 Clause + Patent License.
Please see the contributors guide for a getting started guide and a primer on our procedures and processes.
If you have any questions, feel free to hop in the chat and ask! We welcome anyone of any skill level.
I am now doing a weekly blog segment on development status my personal website. This might be helpful for new contributors to get caught up on what is currently being done.
If you are on github, please hop on over to our gitlab. The github repo is strictly a mirror.
Chat & Support
Our primary chat is on Matrix.
This project and its continued development are made possible by our Hall of Fame members.