What is it?

The Ajira project started as a research project at the Department of Computer Science of the Vrije Universiteit Amsterdam to research a new computing paradigm to efficiently process large amounts of data.

It is inspired by MapReduce (and more specifically by Hadoop). Its ambition is to go beyond it by allowing a much more complex processing on the data that can be performed not only using large computing clusters but also with more modest machines. We call it a Data Engine, because its goal is to make your life easier if you need to process (lots) of data. We hope you'll like it.

The Ajira project was supported by the Dutch national program COMMIT.

Main features

• Automatic parallelization and distribution of the computation. We spend many efforts in trying to hide to the user all the parallelization necessary to operate on large amounts of data. Within one machine, the framework spawns multiple threads. If more machines are used, then it takes care of the nodes communication and ensures correct termination.
• Ability to perform custom operations on the data. In MapReduce only a Map and a Reduce can be executed in sequence. In Ajira, we try to break this mold allowing the user to define an arbitrary number of operations to be executed in whatever sequence is needed.
• Allows different computation flows at runtime. Sometimes you just don't know what you need to execute on your data. In Ajira, you can decide at runtime which actions to take. Multiple actions can be executed in parallel, on the same or different parts of the input. The output of such actions can be merged or be further split. This allows the execution of much more complex programs than a simple Map/Reduce.
• Works on single machines, computing clusters, or even on clouds! In Ajira we use the Ibis IPL library to communicate between machines. The IPL library abstracts the physical location of machines in a logical network. This means that one program can be physically executed using machines in a WAN network, or using machines with heterogeneous connectivity.
• Open source. Ajira is entirely written in Java. All the code is released under the Apache license and can therefore be used or extended by anyone.

Want to know more?

Please, read the documentation that we provide. You can start from the getting started guide, or read our tutorial to learn how to develop with it.

Are you wondering where the name Ajira comes from? You can read it here.