Maintaining reference graphs of globally accessible objects in fully decentralized distributed systems

ABSTRACT

Since the advent of electronic computing, the processors' clock speed has risen tremendously. Now that energy efficiency requirements have stopped that trend, the number of processing cores per machine started to rise. In near future, these cores will become more specialized, and their inter-connections will form complex networks, both on-chip and beyond. This trend opens new fields of applications for high performance computing: Heterogeneous architectures offer different functionalities and thus support a wider range of applications. The increased compute power of these systems allows more complex simulations and numerical computations. Falling costs enable even small companies to invest in multi-core systems and clusters. However, the growing complexity might impede this growth. Imagine a cluster of thousands of interconnected heterogeneous processor cores. A software developer will need a deep knowledge about the underlying infrastructure as well as the data and communication dependencies in her application to partition it optimally across the available cores. Moreover, a predetermined partitioning scheme cannot reflect failing processors or additionally provided resources. In our poster, we introduce J-Cell, a project that aims at simplifying high performance distributed computing. J-Cell offers a single system image, which allows applications to run transparently on heterogeneous multi-core machines. It distributes code, objects and threads onto the compute resources which may be added or removed at run-time. This dynamic property leads to an ad-hoc network of processors and cores. In this network, a fully decentralized object localization and retrieval algorithm guarantees the access to distributed shared objects.

REFERENCES

• 1: T. Fuhrmann. Scalable routing for networked sensors and actuators. In Proc. 2nd Annual IEEE Communications Society Conference on Sensor and Ad Hoc Communications and Networks, Sept. 2005.
• 2: M. Kellerer. Extension of the 'scalable source routing' protocol for addressing and migrating 'global accessible objects'. Diploma thesis, Self-Organizing Systems Group, Technical University of Munich, Germany, Dec. 2008.
• 3: C. Koller. Software transactional memory for the ambicomp distributed java virtual machine. Diploma thesis, System Architecture Group, University of Karlsruhe, Germany, Feb. 2009.
• 4: Hongzhou Liu , Tom Roeder , Kevin Walsh , Rimon Barr , Emin Gün Sirer, Design and implementation of a single system image operating system for ad hoc networks, Proceedings of the 3rd international conference on Mobile systems, applications, and services, June 06-08, 2005, Seattle, Washington [doi>10.1145/1067170.1067187]
• 5: W. Zhu, W. Fang, C.-L. Wang, and F. C. M. Lau. High-Performance Computing on Clusters: The Distributed JVM Approach, chapter 23. 2004.

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.2 COMPUTER-COMMUNICATION NETWORKS
      C.2.4 Distributed Systems
          Subjects: Distributed applications

Additional Classification:
  C. Computer Systems Organization
  C.2 COMPUTER-COMMUNICATION NETWORKS
      C.2.4 Distributed Systems
          Subjects: Network operating systems

General Terms:
Design

Keywords:
globally accessible objects, object migration, single system image