Skip to content

A C++ vectorized database acceleration library aimed to optimizing query engines and data processing systems.

License

Notifications You must be signed in to change notification settings

jaystarshot/velox

 
 

Repository files navigation

Velox logo

Velox is a C++ database acceleration library which provides reusable, extensible, and high-performance data processing components. These components can be reused to build compute engines focused on different analytical workloads, including batch, interactive, stream processing, and AI/ML. Velox was created by Meta and it is currently developed in partnership with IBM/Ahana, Intel, Voltron Data, Microsoft, ByteDance and many other companies.

In common usage scenarios, Velox takes a fully optimized query plan as input and performs the described computation. Considering Velox does not provide a SQL parser, a dataframe layer, or a query optimizer, it is usually not meant to be used directly by end-users; rather, it is mostly used by developers integrating and optimizing their compute engines.

Velox provides the following high-level components:

  • Type: a generic typing system that supports scalar, complex, and nested types, such as structs, maps, arrays, tensors, etc.
  • Vector: an Arrow-compatible columnar memory layout module, which provides multiple encodings, such as Flat, Dictionary, Constant, Sequence/RLE, and Bias, in addition to a lazy materialization pattern and support for out-of-order writes.
  • Expression Eval: a fully vectorized expression evaluation engine that allows expressions to be efficiently executed on top of Vector/Arrow encoded data.
  • Function Packages: sets of vectorized function implementations following the Presto and Spark semantic.
  • Operators: implementation of common data processing operators such as scans, projection, filtering, groupBy, orderBy, shuffle, hash join, unnest, and more.
  • I/O: a generic connector interface that allows different file formats (ORC/DWRF and Parquet) and storage adapters (S3, HDFS, local files) to be used.
  • Network Serializers: an interface where different wire protocols can be implemented, used for network communication, supporting PrestoPage and Spark's UnsafeRow.
  • Resource Management: a collection of primitives for handling computational resources, such as memory arenas and buffer management, tasks, drivers, and thread pools for CPU and thread execution, spilling, and caching.

Velox is extensible and allows developers to define their own engine-specific specializations, including:

  1. Custom types
  2. Simple and vectorized functions
  3. Aggregate functions
  4. Operators
  5. File formats
  6. Storage adapters
  7. Network serializers

Examples

Examples of extensibility and integration with different component APIs can be found here

Documentation

Developer guides detailing many aspects of the library, in addition to the list of available functions can be found here.

Blog posts are available here.

Getting Started

Get the Velox Source

git clone https://github.com/facebookincubator/velox.git
cd velox

Once Velox is checked out, the first step is to install the dependencies. Details on the dependencies and how Velox manages some of them for you can be found here.

Velox also provides the following scripts to help developers setup and install Velox dependencies for a given platform.

Setting up dependencies

The following setup scripts use the DEPENDENCY_DIR environment variable to set the location to download and build packages. This defaults to deps-download in the current working directory.

Use INSTALL_PREFIX to set the install directory of the packages. This defaults to deps-install in the current working directory on macOS and to the default install location (eg. /usr/local) on linux. Using the default install location /usr/local on macOS is discouraged since this location is used by certain Homebrew versions.

Manually add the INSTALL_PREFIX value in the IDE or bash environment, say export INSTALL_PREFIX=/Users/$USERNAME/velox/deps-install to ~/.zshrc so that subsequent Velox builds can use the installed packages.

You can reuse DEPENDENCY_INSTALL and INSTALL_PREFIX for Velox clients such as Prestissimo by specifying a common shared directory.`

Setting up on macOS

On a macOS machine (either Intel or Apple silicon) you can setup and then build like so:

$ ./scripts/setup-macos.sh
$ make

With macOS 14.4 and XCode 15.3 where m4 is missing, you can either

  1. install m4 via brew:
$ brew install m4
$ export PATH=/opt/homebrew/opt/m4/bin:$PATH
  1. or use gm4 instead:
$ M4=/usr/bin/gm4 make

Setting up on Ubuntu (20.04 or later)

The supported architectures are x86_64 (avx, sse), and AArch64 (apple-m1+crc, neoverse-n1). You can build like so:

$ ./scripts/setup-ubuntu.sh
$ make

Setting up on Centos 9 Stream with adapters

Velox adapters include file-systems such as AWS S3, Google Cloud Storage, and Azure Blob File System. These adapters require installation of additional libraries. Once you have checked out Velox, you can setup and build like so:

$ ./scripts/setup-centos9.sh
$ ./scripts/setup-adapters.sh
$ make

Note that setup-adapters.sh supports macOS and Ubuntu 20.04 or later.

Using Clang on Linux

Clang 15 can be additionally installed during the setup step for Ubuntu 22.04/24.04 and CentOS 9 by setting the USE_CLANG environment variable prior to running the platform specific setup script.

$ export USE_CLANG=true

This will install and use Clang 15 to build the dependencies instead of using the default GCC compiler.

Once completed, and before running any make command, set the compiler to be used:

$ export CC=/usr/bin/clang-15
$ export CXX=/usr/bin/clang++-15
$ make

Building Velox

Run make in the root directory to compile the sources. For development, use make debug to build a non-optimized debug version, or make release to build an optimized version. Use make unittest to build and run tests.

Note that,

  • Velox requires a compiler at the minimum GCC 11.0 or Clang 15.0.
  • Velox requires the CPU to support instruction sets:
    • bmi
    • bmi2
    • f16c
  • Velox tries to use the following (or equivalent) instruction sets where available:
    • On Intel CPUs
      • avx
      • avx2
      • sse
    • On ARM
      • Neon
      • Neon64

Build metrics for Velox are published at https://facebookincubator.github.io/velox/bm-report/

Building Velox with docker-compose

If you don't want to install the system dependencies required to build Velox, you can also build and run tests for Velox on a docker container using docker-compose. Use the following commands:

$ docker-compose build ubuntu-cpp
$ docker-compose run --rm ubuntu-cpp

If you want to increase or decrease the number of threads used when building Velox you can override the NUM_THREADS environment variable by doing:

$ docker-compose run -e NUM_THREADS=<NUM_THREADS_TO_USE> --rm ubuntu-cpp

Contributing

Check our contributing guide to learn about how to contribute to the project.

Community

Velox's technical governance mechanics is described in this document.. Components and maintainers are listed here.

The main communication channel with the Velox OSS community is through the the Velox-OSS Slack workspace. Please reach out to [email protected] to get access to Velox Slack Channel.

License

Velox is licensed under the Apache 2.0 License. A copy of the license can be found here.

About

A C++ vectorized database acceleration library aimed to optimizing query engines and data processing systems.

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 96.2%
  • CMake 1.2%
  • Cuda 0.9%
  • C 0.6%
  • Python 0.6%
  • MDX 0.2%
  • Other 0.3%