Trends in Betriebssystemen und Middleware (Forschungsseminar) (Sommersemester 2018)

Dozent: Prof. Dr. Andreas Polze (Betriebssysteme und Middleware)
Website zum Kurs: https://www.dcl.hpi.uni-potsdam.de/meetings/

Allgemeine Information

• Semesterwochenstunden: 2
• ECTS: 3
• Benotet: Ja
• Einschreibefrist: 20.04.2018
• Lehrform: Seminar
• Belegungsart: Wahlpflichtmodul
• Maximale Teilnehmerzahl: 20

Studiengänge, Modulgruppen & Module

Beschreibung

The research seminar "Trends in Operating Systems" is targeted at students who are planning to do their Master's project or Master's thesis with the Operating Systems and Middleware group. We also invite Master's students who are generally interested in OS research.

Students can enroll in the seminar and earn 3 ECTS credits. In order to do so, a prospective candidate has to study recent publications (~3-5) in a topic area. The candidate then has to deliver a presentation in our seminar and hand in a report. Regular attendance is required in order to earn credits.

You report should consist of roughly 8-12 pages LNCS or 4-6 pages IEEE.

Topics

The following is a list of possible topics, mainly inspired by the current research directions at the chair.
However, feel free to suggest cross-cutting and further topics.

• Dependability of software systems
  • Assessment
  • Fault tolerance of service-oriented architectures
  • Fault injection in distributed systems
• Microservice architectures
• Decentralized architectures
• Internet of Things (e.g., Rail2X)
• Machine learning
• Dynamic workload management in distributed systems
• Dynamic, workload-dependent resource scaling (dynamic LPARs)
• Memory optimizations for virtual machines (e.g., memory migration benchmarks)
• Heterogeneous hardware/accelerators: integrated GPUs, APUs, workload offloading, latency-critical workloads
• Programming languages & frameworks: SNAP, Radeon Open Comute, C++17
• PGASUS (PGAS framework): applications, case studies, benchmarks
• Architectures of future server systems
• Memory-centric architectures
• Lock-free data structures
• Embedded systems
• Resource-constrained systems
• Introspection and network tracing
• Log analysis, prediction of maintenance and load
• Multicore/NUMA
• EDCSearch: Optimization of a graph-based algorithm
• Evaluation of cost models for hierachical parallel systems
• Comparison of memory allocation strategies
• Stream stores vs. coherent stores in vectorized applications
• Advanced topology discovery with bandwith and latency measurements: find shared interconnects
• Optimization strategies for matrix multiplication algorithms on business class servers
• Linux kernel experiments
• Windows fault injection experiments
• Formal verification of lock-free code

Leistungserfassung

• Presentation: 30-45min
• Hand in slides and report
• Talk to the supervisor prior to the presentation
  • At least three weeks prior to the presentation: define directions and papers to read
  • At least one week prior to the presentation: discuss quality of slides
• Be prepared for a Q&A (discussions, backup slides, ...)
• Attendance is expected

Termine

Tuesday 11.00, C-1.x

Zurück