• Stars
    star
    1,429
  • Rank 31,618 (Top 0.7 %)
  • Language
    Rust
  • License
    Apache License 2.0
  • Created over 3 years ago
  • Updated almost 2 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

TensorBase is a new big data warehousing with modern efforts.

chat on Discord

Status of the Project

TensorBase hasn't been updated for a while. Thanks for all friends' concern and inquiries, we reply as follows:

TensorBase hopes the open source not become a copy game. TensorBase has a clear-cut opposition to fork communities, repeat wheels, or hack traffics for so-called reputations (like Github stars). After thoughts, we decided to temporarily leave the general data warehousing field.

Here, let's recap all the world's first of TensorBase:

  1. The world's first ClickHouse compatible open-source implementation.
  2. 2x faster write throughput than that of ClickHouse (based on our bug fixed Rust client, you can get ~1.7x speedup by our another simple concurrent script here).
  3. Faster query speed in the simple aggregation than that of ClickHouse(benchmarked against ClickHouse 2021.6).
  4. First no-LSM, write and read optimized storage layer proposed.
  5. First make "copy-free, lock-free, async-free, dyn-free" happened in an open-source DBMS's critical path.
  6. First DBMS running on the real-world RISC-V hardware.
  7. First top-performance whole-lifecycle JIT SQL query engine(not completely open sourced, but we release the initial prototype in the network which you can try, and more you can get the ideas from the blogs, presentations and videos.)...

For people looking for production level data warehouse solution, we still recommend ClickHouse. We wish that ClickHouse can learn from these work and evolve itself to better.

For people who want to learn how a database system can be built up, or how to apply modern Rust to the high performance field, or embed a lightweight data analysis system into your own big one. You can still try, ask or contribute to TensorBase. The committers are still around the community. We will help you in all kinds of interesting things pursued in the project by us and maybe you. We still maintain the project to look forward to meeting more database geniuses in this world, although no new feature will be added in the near future.

The core team of TensorBase has moved to another new type of domain-specific database. We are hiring!


What is TensorBase

TensorBase is a new big data warehousing with modern efforts.

TensorBase is building on top of Rust, Apache Arrow and Arrow DataFusion.

TensorBase hopes to change the status quo of bigdata system as follows:

  • low efficiency (in the name of 'scalable')
  • hard to use (for end users) and understand (for developers)
  • not evolving with modern infrastructures (OS, hardware, engineering...)

Features

  • Out-of-the-box to play ( get started just now )
  • Lighting fast architectural performance in Rust ( real-world benchmarks )
  • Modern redesigned columnar storage
  • Top performance network transport server
  • ClickHouse compatible syntax
  • Green installation with DBA-Free ops
  • Reliability and high availability (WIP)
  • Cluster (WIP)
  • Cloud-Native Adaptation (WIP)
  • Arrow dataLake (...)

Architecture (in 10,000 meters altitude)

arch_base

Quick Start

play_out_of_the_box

Benchmarks

TensorBase is lighting fast. TensorBase has shown better performance than that of ClickHouse in simple aggregation query on 1.47-billion rows NYC Taxi Dataset.

TensorBase has enabled full workflow for TPC-H benchmarks from data ingestion to query.

More detail about all benchmarks seen in benchmarks.

Roadmap

Community Newsletters

Working Groups

Working Group - Engineering

This is a wg for engineering related topics, like codes or features.

Working Group - Database

This is a higher kind wg for database related topics, like ideas from papers.

Join these working groups on the Discussions or on Discord server.

Communications

Wechat group or other more are on community

Contributing

We have a great contributing guide in the Contributing.

Documents (WIP)

More documents will be prepared soon.

Read the Documents.

License

TensorBase is distributed under the terms of the Apache License (Version 2.0), which is a commercial-friendly open source license.

It is greatly appreciated that,

  • you could give this project a star, if you think these got from TensorBase are helpful.
  • you could indicate yourself in Who is Using TensorBase, if you are using TensorBase in any project, product or service.
  • you could contribute your changes back to TensorBase, if you want your changes could be helpful for more people.

Your encouragements and helps can make more people realize the value of the project, and motivate the developers and contributors of TensorBase to move forward.

See LICENSE for details.