ROME 2019 - Workshop on Runtime, Operating Systems and Middleware towards Exascale (ROME)

High-performance computing itself is a well-established scientific domain. However, to pave the way towards exascale, further evolutionary steps are currently required to be taken by moving away from common multi-/many-core clusters towards novel and innovative heterogeneous architectures. Such systems, equipped with a significantly higher number of (heterogeneous) cores than today’s supercomputers, pose challenges in both hardware and software design. On the hardware side, new processor and accelerator architectures, complex on-chip networks, deep memory-hierarchies and interconnect technologies will enrich the respective research areas. However, in keeping with its tradition, the ROME workshop focuses on the software side because without complying system software, runtime and operating system support, all these new hardware facilities cannot be exploited. Hence, the challenges in hardware/software co-design are to step beyond traditional approaches and to venture new strategies for runtime, middleware and operating system designs in order to exploit the theoretically available performance of upcoming hardware features as effectively and energy-consciously as possible. For gaining the required power-efficiency, scalability and manageability an exascale system will demand for, both virtualization as well as machine-oriented optimization will have to be exploited jointly. For doing so, customized operating systems, hypervisors and unikernels will be required to leverage an efficient employment of virtualization and an effective machine optimization on a broader scale – which has of course to be supported and enhanced by a corresponding runtime and middleware.
Topics and Call for Papers
Authors from all related disciplines are invited to submit unpublished papers regarding their work on software research regarding operating systems and runtime environments in the domain of high-performance and parallel computing. The topics of interest include, but are not limited to:
New approaches for operating systems on novel (heterogeneous) architectures
Lightweight/specialized operating systems, unikernels and hypervisors
Virtualization techniques (e. g. containers) for high-performance computing
Management, deployment of virtualized environments and orchestration
Extensions for operating systems to support emerging memory technologies
System software for enabling parallelism at an extreme scale
Bare-metal programming and system software for massive parallel systems
Many-core/multi-node aware runtime support for large-scale applications
Message-passing interfaces and middleware for extreme scale computing
Heterogeneity-, modularity- and/or hierarchy-aware middleware for HPC
Software stacks for new concepts of compute acceleration by GPUs, FPGAs and many-core architecture
Middleware for performance and power analysis on massive parallel systems
Runtime support and kernel extensions for power-aware high-performance computing
System noise analysis and prevention for massive parallel systems