Kolloquium des Studiendekanats Elektrotechnik, Informatik und Mathematik (EIM)
Das Studiendekanat Elektrotechnik, Informatik und Mathematik der Technischen Universität Hamburg (TUHH) freut sich, im Rahmen seines Kolloquiums eine weitere Antrittsvorlesung und einen Forschungsvortrag aus dem Bereich der Betriebssysteme und System-Software mit der gesamten TUHH und mit der Öffentlichkeit zu teilen. Dieses Kolloquium des Dekanats EIM beginnt am
Donnerstag, den 7. April 2022 ab 14:00 Uhr
mit folgendem kurzen Programm:
14:00 Uhr: Antrittsvorlesung „New Directions for Managing Memory“. Prof. Christian Dietrich, Operating System Group, Technische Universität Hamburg (TUHH)
14:45 Uhr: A (System-) Software Perspective on Disruptive Hardware Technologies. Prof. Olaf Spinczyk, Arbeitsgruppe Eingebettete Softwaresysteme, Universität Osnabrück
Traditionally, memory is the scarce resource that operating systems virtualize for their users. While CPU time replenishes at a one-to-one rate every second, memory is not consumable and we can assign each page frame and each disk block to exactly one purpose at a time. In order to still maintain the illusion of abundant and persistent memory, operating systems provide virtual-memory abstractions, file systems, heuristic data-transfer policies that all require complex bookkeeping structures and put the operating system into the position of an active intermediate into the hot data path. However, current hardware trends, like ultra-fast NVMe SSDs and non-volatile RAM, force us to rethink operating system-mediated management. We no longer have to manage scarcity, but we have to swim in the new abundance without drowning.
Christian Dietrich leads the Operating System Group at TUHH since April 2021, where he and his team investigate on new operating system abstractions and more efficient mechanisms to assist users in exploiting current and future hardware trends. His main research topics include:
Virtual memory as the most efficient interface that operating systems provide.
Dynamic variability that allows applications to concurrently use and replace software components without interrupting service quality.
Resilient systems that remain operational even in the presence of necessary updates, security attacks, and transient hardware faults.
In his inaugural lecture, Christian Dietrich will present three ongoing research projects that center around the topic of memory management but are closely related to his other research interests. With Explicit File-mappings, we give data-intensive applications full control about the data placement without requiring a re-design of the system. Morsels are a unifying abstraction to procure, share, and transfer large amounts of (non-volatile) memory within a (potentially disaggregated) system. And with MMViews, users become able to maintain multiple, synchronized versions of their address space within a single process. In all of these projects, the Operating Systems Group questions long-standing abstractions and operates with the mission statement:
Nobody wants an operating system, only its services are required.
A (System-) Software Perspective on Disruptive Hardware Technologies
Over the past decade the architecture of modern computer hardware has changed dramatically. While today’s applications and operating systems are still struggling to efficiently and correctly handle multi-core CPUs, the development in hardware technology has already taken next major leaps. Software technology is again falling behind. Examples for open questions are:
How to efficiently use non-uniform and disaggregated memory architectures?
How to make use of ordinary CPU cores, integrated FPGA, GPGPUs, processing-in-memory (PIM), and AI accelerators in a coordinated manner?
How to integrate persistent memory modules into our software systems?
How to handle system-wide properties such as energy management or techniques to deal with aging of hardware components?
In the research project “MxKernel: A Bare-Metal Runtime System for Database Operations on Heterogeneous Many-Core Hardware” we have started out to tackle some of the aforementioned problems on the system software level. The talk will cover the motivation of this work, the proposed system software architecture, which is designed from scratch, and the results that we have already obtained.
However, it turned out that there are too many challenges for a single project to handle. Therefore, we have proposed the DFG priority program “Disruptive Memory Technologies”, which will start in summer 2022. The talk will also give an outlook on this program, which will be a coordinated effort to explore methods for efficient use of memory technologies that affect software development on multiple layers of the hardware/software stack.
Olaf Spinczyk leads the Embedded Software Systems Group at Osnabrück University since 2018. During his academic career he worked at the University of Magdeburg, FAU Erlangen-Nuremberg, and from 2007 to 2018 as a professor an Embedded System Software at TU Dortmund. His research covers a broad spectrum of topics in the context of system software construction with special focus on future embedded systems. Olaf is highly active in the German Computer Science Society (GI) and serves as speaker of the new DFG priority program 2377 “Disruptive Memory Technologies”.