Traditional databases are separated into ones for current data from the day-to- day business processes and ones for reporting and analytics. For fast moving businesses moving data from one silo to another is cumbersome and takes too much time. As a result the new data arriving in the reporting system is already old by the time it is loaded. HYRISE proposes a new way to solve this problem: It analyzes the query input and reorganizes the stored data in different dimensions.
In detail, HYRISE partitions the layout of the underlying tables in a vertical and horizontal manner depending on the input to this layout management component. The workload is specified as a set of queries and weights and is processed by calculating the layout dependent costs for those queries. Based on our cost-model we can now calculate the best set of partitions for this input workload. This optimization allows great speed improvements compared to traditional storage models.
This database provides the implementation to the above mentioned ideas.
This system is a research prototype and by no means meant to be used in production use and we cannot guarantee that it will work in such environments. The goal of an early version of HYRISE was to show that for in-memory databases the flexibility to choose from different physical layouts is of great importance. In addition, this system should provide the necessary framework for teaching advanced database techniques to students in an undergraduate and graduate level.
Our goal is to improve the performance and the set of features to support more and more capabilities that you would expect from a traditional relational database. But no promises, when this will happen.
The following features are currently implemented (not always optimal and convenient):
- Database loading from CSV, HYRISE text files, binary dumps
- Most important database operators and indexes including re-compression and dump to disk
- Automatic layout calculation and layout switch as front-end accessible plan operations
All plan operations are currently hand-coded in JSON an then queried against the database. SQL compilation is currently not available but somewhere on the road-map.
Master status:
Supported (actively tested through CI and in development) systems:
- Ubuntu (>13.04)
Likely supported:
- MacOSX 10.7
HYRISE expects a C++11-capable compiler. We recommend g++ >=4.7, testing is done against g++ 4.7, g++ 4.8 and clang 3.2.
You can use the official HYRISE development image to start a docker container with all dependencies for building HYRISE already installed. See Docker README for more information on how to use it.
Using VirtualBox and Vagrant, you can automatically setup a HYRISE development environment. (This is the recommended way of getting started with HYRISE development).
Check out this repository to easily setup a development VM.
It should be sufficient to run the install_$version.sh script from the tools/ folder, where version corresponds to a Ubuntu version.
First check if all submodules are initialized correctly:
git submodule update --init
To start the build you will only need to execute a simple command line command to start. The given make file provides all necessary targets.
A simple build is executed by
make
To build and execute the test suites using predefined tables issue
make test
The build can be configured through a build settings file settings.mk.
Copy the file settings.mk.default to settings.mk and if necessary
modify settings. The most prominent one is PRODUCTION to
disable the debug build and HYRISE_ALLOCATOR for using tcmalloc or
jemalloc instead of the libc allocator.
To enable an optimized production build with g++-4.8, a settings.mk file
may look as follows:
COMPILER := g++48
PRODUCTION := 1
When the settings file has been changed, running make should result in a
complete rebuild of HYRISE.
More options are discussed in the documentation.
The project provides some documentation about how to develop and use HYRISE. The documentation can be build using:
make docs
The documentation contains information about how to use HYRISE and how to implement own plan operations and table types etc.
If you have any questions feel free to post on the developer mailing list:
Pull requests and issues reports are always welcome. The easiest way to participate is to clone HYRISE and submit patches that we can merge including tests.
If you have any questions feel free to contact the maintainers
- David Schwalb (@schwald)
- Martin Faust (@martinfaust)
The following people contributed to HYRISE in various forms listed in alphabetical order :)
- Alexander Franke
- Christian Tinnefeld
- Clemens Frahnow
- David Eickhoff
- David Schwalb
- Friedhelm Filler
- Georg Hoefer
- Henning Lohse
- Holger Pirk
- Jan Kossmann
- Jan Oberst
- Jens Krueger
- Johannes Wust
- Jonas Witt
- Kai Höwelmeyer
- Karsten Tausche
- Marco Hornung
- Markus Dreseler
- Martin Boissier
- Martin Faust
- Martin Grund
- Martin Linkhorst
- Marvin Keller
- Marvin Killing
- Matthias Lleine
- Maximilian Schneider
- Pedro Flemming
- Robert Strobl
- Sebastian Blessing
- Sebastian Hillig
- Sebastian Klose
- Tim Berning
- Tim Zimmermann
- Uwe Hartmann
HYRISE is licensed as open source after the OpenSource "Licence of the Hasso-Plattner Institute" declared in the LICENSE file of this project.
The reasoning stems from German copyright law. Common BSD and MIT licenses are not necessarily compatible with liability issues in German copyright legislation. Thus, HYRISE uses an specifically designed open source license deemed compatible with German law. The most prominent difference between our license and commonly used BSD or MIT is the inclusion of certain liabilities as dictated by German copyright law.