Section: Overall Objectives

Summary of the main contributions of the team in 2009

During this year, RESO team had main contributions in the following fields:

Axis 1 : Protocol implementations and networking equipments

• Exploration of the data plane virtualization cost and opportunities in software routers (VXRouter);

• Design of an Autonomic Network Programming Interface

• High availability for clustered stateful network equipments

• Models and software frameworks for energy efficiency in large scale distributed systems

Axis 2 : End-to-end Quality of Service and Transport layer

• Optimization algorithms for network resource sharing in very high speed networks (BDTS).

• Algorithms for dynamic bandwidth provisioning based on flows scheduling and aggregation

• Design of a language for specifying virtual infrastructures (VXDL)

• Study on network and system virtualisation for virtual private infrastructure creation and usage.

• Distributed algorithms for bandwidth sharing in mobile or very high speed environments.

• Analysis of opportunity of extending the size of transfer units (XLFrames).

• Analysis of flow scheduling and sampling-based scheduling

• Analysis of flow-based routing

Axis 3: High Speed Network's traffic metrology and statistical analysis

• Derivation of a maximum likelihood estimator of heavy tail distribution index from censored data. This estimator was applied to a set of incomplete flows reconstructed from a packet sub-sampled high speed network traffic, to get the tail index of the corresponding flow size distribution.

• Extension of Taqqu's relation relating heavy-tailed flow size distributions to long range dependence, to take into account the correlation between flow size and flow throughput.

• Refine the relations between observed statistical scaling properties of Internet traffic and performance measures (delay, loss rate) as QoS metrics.

• Identification on a long-lived TCP flow, of a new type of scale invariance property with multifractal support. Demonstration of a (almost sure) large deviation principle guaranteeing the identifiability of the corresponding multifractal spectrum from a finite size real trace. Adaptation to TCP markovian models to characterize fairness and sources' synchronization.

Axis 4: Network Services for high demanding applications

• Design of a network service for dynamic bandwidth provisioning in very high speed environments;

• Specification of a network resource scheduling, virtualization and reconfiguration component in a service oriented approach (Carriocas)

• Design and development of MPI5000, a communication layer for MPI over wide area network

• Traffic monitoring of LCG (CERN LHC Grid) 10Gb/s link at packet resolution to characterize Grid traffic;

• Design and development of the HIPerNet software for network-aware virtual cluster management tool.

• Pursue the collaborations for the development and usage of the GRID5000 international optical interconnections to Netherlands (DAS3) and Japan (Naregi) in collaboration with RENATER;

• Design and development of a metrology infrastructure for fine grain traffic monitoring in Grid5000.