In the recent past, there has been important progress in the production-ready implementation of non-volatile memory (NVM). In contrast to today's common dynamic random access memory (DRAM), it does not lose its stored information even if the power supply is interrupted. In addition, unlike conventional non-volatile storage media, data stored in NVM can be processed directly by a computer's processor. This gives NVM performance characteristics which are partially comparable to those of DRAM, e. g. access speed or bandwidth.

However, there are different basic technologies to implement non-volatile main memory. These differ greatly in terms of write and read speed or the available capacity, for example.

Certain attributes of non-volatile main memories theoretically lead to advantages in the design of computer architectures. For example, the restart time of systems after a power failure could be reduced because loading information stored on durable disks such as hard drives into a computer's main memory becomes unnecessary. In addition, non-volatile main memory can save energy and improve application performance by simplifying certain components in the operating system. However, taking advantage of these benefits is a significant challenge on the software side.

The FOSSIL project focuses on making the advantages of NVM usable for applications. For this purpose, an existing microkernel operating system is changed to support a quick restart of applications, for example, after a power failure. Furthermore, the project aims to analyze to what extent the distribution of application data among different types of memory affects the overall performance of the respective application. The energy saving potential created by non-volatile main storage is also to be researched. Moreover, an essential requirement of the project is the support of legacy programs that are not adapted to the peculiarities of non-volatile memory.

The project is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) within the Priority Program 2377 – Project Number: 502457159.

Project duration: 2022 – 2025

Contact: Dr. Michael Roitzsch (michael.roitzsch@barkhauseninstitut.org), Till Miemietz (till.miemietz@barkhauseninstitut.org)