Submitted by gabriele on
TERAFLUX will be represented at the Workshop / mini-symposium on trends, issues and challenges in future high-end computing. In the context of the All-Hands-Meeting 2010, Cardiff, UK, 13-16 September 2010.
The meeting provides a forum in which information on e-Science projects from all disciplines can be communicated and where the capabilities being developed within projects can be demonstrated.
More information on AHM2010 can be found here.
Registration for AHM 2010 is now open.
A late fee of £50 will apply from 16 August and registration closes on 23 August.
Please see the registration page for further details.
Call for Participation: AHM2010
Workshop / mini-symposium on trends, issues and challenges in future high-end computing. In the context of the All-Hands-Meeting 2010, Cardiff, UK, 13-16 September 2010
With the movement from the “frequency race” to the “core race”, issues formerly restricted primarily to High Performance Computing (namely scalability, distribution etc.) become more and more relevant for day-to-day computing. Future applications can no longer rely on stronger (in the sense of faster) compute infrastructures, but have to cater for more parallelism – this denotes a complete conceptual change for application & tools developers who so far extended the capabilities of their system simply by relying on the fact that future computers will be fast enough for smooth execution.
At the same time, the range of core and computer types constantly increase, leading to a wider heterogeneity of resource infrastructures for which the applications have to cater. Standard monolithic operating systems try to address a variety of platforms on the cost of performance, whilst efficient OS as used for High Performance Computers are typically specifically devised / adapted to the respective infrastructure. Furthermore, the scale will increase beyond the manageability of a single OS instance, and future internet and cloud systems will incorporate multiple concurrent OS instances. Consequently, developers, middleware providers, and resource administrators will face multiple issues when trying to scale and efficiently execute future applications. Implicitly and at the same time the size of data and the range of process types increases nearly exponentially and obviously faster than networks and platforms can deal with. Distributed applications, concurrent access, multi-tenancy etc. all will grow in significance over the next years.
In the scope of this symposium-like workshop, trends in high end computing development shall be discussed in-depth with a specific focus on challenges and issues to be addressed via current and future research. This relates in particular to the issues arising from the move of single, high-frequency cores in distributed environments to the large-scale, widely distributed many-core devices currently under way. In this context, two major strands can be identified that will be pursued by this workshop:
STRAND 1: SCALABLE INFRASTRUCTURES. The increasing scale and heterogeneity of infrastructures (ranging from many-core systems over large-scale clusters to clouds) implies that classical approaches to their management are not sufficient anymore: monolithic OS scale badly, cache coherency takes exponentially more cycles etc. Topics in this strand include, but are not restricted to:
- next generation operating systems
- next generation processor architectures
- management of cache, interconnect etc. in large scale many-core systems
- clouds on a chip
STRAND 2: SCALABLE CODE AND DATA. In order actually make use of the infrastructure, applications and data need to become more scalable, which implies in particular that it becomes easier for the user to develop for and exploit parallel systems. This strand includes, amongst other, issues related to:
- segmentation and distribution methods
- parallel programming models
- efficient scaling behaviour
- High performance data access
- Data replication strategies for large data sets
- Performance optimisation techniques
- Metadata management
- HPC Clouds
Workshop Committee
- Lutz Schubert, High Performance Computing Centre Stuttgart (chair)
- Matthias Assel, HLRS Stuttgart (chair)
- Götz Brasche, Europäisches Microsoft Innovations Center GmbH
- Johannes Helander, Europäisches Microsoft Innovations Center GmbH
- Rui Aguiar, Instituto de Telecomunicacoes, Universidade de Aveiro
- João Paulo Barraca, Instituto de Telecomunicacoes, Universidade de Aveiro
- Giuseppe Lipari, Scuola Superiore Sant'Anna
- Tommaso Cucinotta, Scuola Superiore Sant'Anna
- Jan Kuper, University of Twente
- Christiaan Baaij, University of Twente
- Roberto Giorgi, Universita' degli Studi di Siena