DXRAM is a high speed, scalable, and reliable key-value store for billions of data items. Its design is focusing on low-latency data access, high throughput, and resource efficiency. DXRAM is implemented in Java, open source, running on all Linux distributions, and ready for the cloud. The DXRAM research project has been partially funded by the German Science Foundation.
Performance. DXRAM provides read and write access in tenths of microseconds by storing all data always in memory. The design is optimized for efficiently managing billions of small data objects. We support Gigabit Ethernet as well as Infiniband.
Reliability is provided by automatic distributed replication and transparent persistence for all data. DXRAM can recover a failed server in a couple of seconds from a SSD-based log ensuring high availability of all stored data.
Costs. A very low memory overhead allows DXRAM to store billions of data items per server. It is is free, provided as open source, implemented in Java and can be run on any Linux distribution and is ready to be deployed in public clouds.
Scalability. DXRAM can be used on a couple of machines or large-scale sets with hundreds of servers depending on data volume or computation power needs.
Prof. Dr. Michael Schöttner (coordinator)
Dr. Kevin Beineke (replication, parallel recovery, network)
Dr. Stefan Nothaas (Infiniband, memory management)
Filip Krakowski (data migration)
Fabian Ruhland (InfiniBand)
- Burak Akguel, MapReduce-Framework and Monitoring
- Maximilian Loose, REST API
- Kai Neyenhuys, Indexing
- Mikel Bahn, Graph Applications
- Nils Axer, Consenus-Service
- Julien Bernhart, Automatic Code Management
- Ruslan Curbanov, DirectAccess, DXDDL, and Graphalytics Cient
- Christian Gesse, DXNet thread-management
- Florian Hucke, Memory-Management
- Lars Mehnert, Distributed Hash-Map
- Sven Gasterstädt, Graph-Formats
- Constantin Eiteneuer, Page-Rank
- Florian Völz, Distributed graph loading
- Dominik Kuhnen, Adbvanced Indexing
- "Low-Latency Data Access in a Java-based Distributed In-Memory Key-Value Storage",
Stefan Nothaas, 2019.
- "Schnelle parallele Fehlererholung in verteilten In-Memory Key-Value Systemen",
Kevin Beineke, 2018.
- "Metadaten-Verwaltung in einem verteilten RAM-basierten Speicherdienst",
Florian Klein, 2015.