YugaByte Database is the open source, transactional, high-performance database for building internet-scale, globally-distributed applications. This repository contains the Community Edition of the YugaByte Database.
Table of Contents
- About YugaByte DB
- Getting Started
- Developing Apps
- Building YugaByte code
- Reporting Issues
About YugaByte DB
Built using a unique combination of log-structured merge document store, auto-sharding, per-shard distributed consensus replication and multi-shard ACID transactions (inspired by Google Spanner), YugaByte DB is world's only distributed database that is both non-relational (supports Redis-compatible key-value & Cassandra-compatible flexible schema APIs) and relational (PostgreSQL-compatible distributed SQL API) at the same time. It is purpose-built to power fast-growing online services on public, private and hybrid clouds with transactional integrity, low latency, high throughput and multi-region scalability while also providing unparalleled data modeling freedom to application architects. Enterprises gain more functional depth and agility without any cloud lock-in when compared to proprietary cloud databases such as Amazon DynamoDB, Microsoft Azure Cosmos DB and Google Cloud Spanner. Enterprises also benefit from stronger data integrity guarantees and higher performance than those offered by legacy open source NoSQL databases such as MongoDB and Apache Cassandra.
YugaByte DB architecture has 2 layers. At the core is DocDB, YugaByte DB's distributed document store. DocDB is the common database engine for the YugaByte DB API layer. Applications interact with the YugaByte DB API layer using one or more of the APIs highlighted below.
YugaByte DB APIs
YugaByte DB supports both Transactional NoSQL and Distributed SQL APIs.
- YugaByte Dictionary Service (YEDIS) - A Redis-compatible Key-Value API with support for hash, sorted sets, pub/sub and time series data structures.
- YugaByte Cloud Query Language (YCQL) - A Cassandra-compatible Flexible Schema API with strong consistency, distributed ACID transactions, globally consistent secondary indexes and a native JSONB data type.
- YugaByte Structured Query Language (YSQL) - A PostgreSQL-compatible Distributed SQL API (currently in beta) with linear write scalability and extreme fault tolerance against infrastructure failures.
For transactional, internet-scale workloads, the question of which API to choose is a trade-off between data modeling richness and query performance. On one end of the spectrum is the YEDIS API that is completely optimized for single key access patterns, has simpler data modeling constructs and provides blazing-fast (sub-ms) query performance. On the other end of the spectrum is the YSQL API that supports complex multi-key relationships (through JOINS and foreign keys) and provides normal (single-digit ms) query performance. This is expected since multiple keys can be located on multiple shards hosted on multiple nodes, resulting in higher latency than a key-value API that accesses only a single key at any time. At the middle of the spectrum is the YCQL API that is still optimized for majority single-key workloads but has richer data modeling features such as globally consistent secondary indexes (powered by distributed ACID transactions) that can accelerate internet-scale application development significantly.
DocDB, YugaByte DB's Distributed Document Store
DocDB builds on top of the popular RocksDB project by transforming RocksDB from a key-value store (with only primitive data types) to a document store (with complex data types). Every key is stored as a separate document in DocDB, irrespective of the API responsible for managing the key. DocDB’s sharding, replication/fault-tolerance and distributed ACID transactions architecture are all based on the the Google Spanner design first published in 2012.
Here are a few resources for getting started with YugaByte DB:
- Quick start guide - install, create a local cluster and read/write from YugaByte DB.
- Explore core features - automatic sharding & re-balancing, linear scalability, fault tolerance, tunable reads etc.
- Ecosystem integrations - integrations with Apache Kafka/KSQL, Apache Spark, JanusGraph, KairosDB, Presto and more.
- Real world apps - sample real-world, end-to-end applications built using YugaByte DB.
- Architecture docs - to understand how YugaByte DB was designed and how it works
Cannot find what you are looking for? Have a question? We love to hear from you - please file a GitHub issue.
Here is a tutorial on implementing a simple Hello World application for YugaByte DB's YCQL and YEDIS APIs in different languages:
We are constantly adding documentation on how to build apps using the client drivers in various languages, as well as the ecosystem integrations we support. Please see our app-development docs for the latest information.
Once again, please post your questions or comments as a GitHub issue if you need something.
Building YugaByte code
Prerequisites for CentOS 7
CentOS 7 is the main recommended development and production platform for YugaByte.
Update packages on your system, install development tools and additional packages:
sudo yum update sudo yum groupinstall -y 'Development Tools' sudo yum install -y ruby perl-Digest epel-release ccache git python2-pip sudo yum install -y cmake3 ctest3
Also we expect
ctest binaries to be at least version 3. On CentOS one way to achive
this is to symlink them into
sudo ln -s /usr/bin/cmake3 /usr/local/bin/cmake sudo ln -s /usr/bin/ctest3 /usr/local/bin/ctest
You could also symlink them into another directory that is on your PATH.
We also use Linuxbrew to provide some of the third-party
dependencies on CentOS. During the build we install Linuxbrew in a separate directory,
~/.linuxbrew-yb-build/linuxbrew-<version>, so that it does not conflict with any other Linuxbrew
installation on your workstation, and does not contain any unnecessary packages that would
interfere with the build.
We don't need to add
~/.linuxbrew-yb-build/linuxbrew-<version>/bin to PATH. The build scripts
will automatically discover this Linuxbrew installation.
Prerequisites for Mac OS X
/usr/bin/ruby -e "$( curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
Install the following packages using Homebrew:
brew install autoconf automake bash bison ccache cmake coreutils flex gnu-tar icu4c libtool maven \ ninja pkg-config pstree wget zlib
Also YugaByte DB build scripts rely on Bash 4. Make sure that
which bash outputs
/usr/local/bin/bash before proceeding. You may need to put
/usr/local/bin as the first directory
on PATH in your
~/.bashrc to achieve that.
Prerequisites for drivers and sample apps
YugaByte DB core is written in C++, but the repository contains Java code needed to run sample applications. To build the Java part, you need:
- JDK 8
- Apache Maven.
Also make sure Maven's
bin directory is added to your PATH, e.g. by adding to your
if you've installed Maven into
For building YugaByte DB Java code, you'll need to install Java and Apache Maven.
YugaByte DB and Apache Cassandra use different approaches to split data between nodes. In order to route client requests to the right server without extra hops, we provide a custom load balancing policy in our modified version of Datastax's Apache Cassandra Java driver.
The latest version of our driver is available on Maven Central. You can build your application using our driver by adding the following Maven dependency to your application:
<dependency> <groupId>com.yugabyte</groupId> <artifactId>cassandra-driver-core</artifactId> <version>3.2.0-yb-18</version> </dependency>
Building the code
Assuming this repository is checked out in
~/code/yugabyte-db, do the following:
cd ~/code/yugabyte-db ./yb_build.sh release --with-assembly
The above command will build the release configuration, put the C++ binaries in
build/release-gcc-dynamic-community, and will also create the
build/latest symlink to that
directory. Then it will build the Java code as well. The
--with-assembly flag tells the build
script to build the
yb-sample-apps.jar file containing sample Java apps.
For Linux it will first make sure our custom Linuxbrew distribution is installed into
Running the C++ tests
To run all the C++ tests you can use following command:
./yb_build.sh release --ctest
If you omit
release argument, it will run java tests against debug YugaByte build.
To run specific test:
./yb_build.sh release --cxx-test util_monotime-test
Also you can run specific sub-test:
./yb_build.sh release --cxx-test util_monotime-test --gtest_filter TestMonoTime.TestCondition
Building Java code alone
You can skip building C++ code, this can be useful when you only need to rebuild Java code:
cd ~/code/yugabyte-db ./yb_build.sh --scb
Running the Java tests
Given that you've already built C++ and Java code you can run Java tests using following command:
./yb_build.sh release --scb --sj --java-tests
If you omit
release argument, it will run java tests against debug YugaByte build, so you should then either
build debug binaries with
./yb_build.sh or omit
--scb and then it will build debug binaries automatically.
Alternatively, to run specific test:
./yb_build.sh release --scb --sj --java-test org.yb.client.TestYBClient
To run a specific Java sub-test within a test file use the # syntax, for example:
./yb_build.sh release --scb --sj --java-test org.yb.client.TestYBClient#testClientCreateDestroy
Viewing log outputs of Java tests
You can find Java tests output in corresponding directory (you might
need to change
yb-client to respective Java tests module):
$ ls -1 java/yb-client/target/surefire-reports/ TEST-org.yb.client.TestYBClient.xml org.yb.client.TestYBClient-output.txt org.yb.client.TestYBClient.testAffinitizedLeaders.stderr.txt org.yb.client.TestYBClient.testAffinitizedLeaders.stdout.txt … org.yb.client.TestYBClient.testWaitForLoadBalance.stderr.txt org.yb.client.TestYBClient.testWaitForLoadBalance.stdout.txt org.yb.client.TestYBClient.txt
Note that the YB logs are contained in the output file now.
YugaByte DB Community Edition is distributed under an Apache 2.0 license. See the LICENSE.txt file for details.