Innovation for Every Engineer.

A better data storage in the making

Techniques and coding theories in information sciences have helped to design new concepts for faster and better data storage. Rapid and reliable data storage is crucial for the success of businesses, online platforms, information repositories and research centres. The EU-funded INFO-STORE (Novel data storage by advancing information sciences) project investigated how to overcome challenges and limitations in emerging data storage applications.Combining theory, practice and expert discussions in a series of five subprojects, the project examined how to improve information storage through breakthroughs in coding theory and related theoretical disciplines.

The project made significant progress with respect to multilevel non-volatile memories and their trade-offs between storage density and read/write speeds. This included designing new coding frameworks to improve read/write speeds, maximising the number of writes, reading memory cells in parallel, and balancing density and speed in multilevel memories.

To develop high-throughput memory systems for network switches and routers, INFO-STORE produced a coding framework that switches packets between input and output ports. This also involved new codes for optimal switching guarantees, algorithms that exploit such codes in a network switch or router, and compression of forwarding databases. The achievements in this vein overcame the challenge of memory contention in network switches and routers.

High-density memristor (resistive memory) storage and associated reliability challenges were also studied to avoid a key obstacle known as sneak paths. In addition, work on memristor issues involved using regular arrays of memristor devices that combine storage and logic.

Another important project milestone involved distributed storage systems and the need for flexible data reconstruction in the presence of failing and busy nodes. To overcome this obstacle, INFO-STORE developed a 2D coding framework that allows only a number of nodes to provide a small portion of their stored information – i.e. partial delivery.

Lastly, researchers proposed efficient data distribution schemes that work well with the short packet blocks. As data distribution schemes using fountain codes operate optimally only when the packet blocks are very large, the team produced a coding scheme to optimise encoder operation for the receiver's state. This lowers the overhead considerably for short blocks.

The project's results were followed by many other ideas and proposals, including a publication on elastic compression of content in web servers to improve server security against denial-of-service attacks. The project's research findings have already begun contributing to the development of quicker, better and safer data storage.




Related Posts