Pascale Vicat-Blanc Primet

Scheduling deadline-constrained bulk data transfers to minimize network congestion

Tight coordination of resource allocation among end points in Grid networks often requires a data mover service to transfer a voluminous dataset from one site to another in a specified time interval. With flexibility at its best, the transfer can start from any time after its arrival, use any and even time variant bandwidth value, as long as it is completed before its deadline. Given a set of such tasks, we study the Bulk Data Transfer Scheduling (BDTS) problem, which searches for the optimal bandwidth allocation profile for each task to minimize the overall network congestion. We show that the multi-interval scheduling, which divides the active window of a task into multiple intervals and assigns bandwidth value independently in each of them, is both sufficient and necessary to attain the optimality in BDTS. Specifically, we show that BDTS can be solved in polynomial time as a Maximum Concurrent Flow Problem. The optimal solution attained is in the form of multi-interval scheduling with the number of intervals upper-bounded. Simulations are conducted over several representative topologies to demonstrate the significant advantage of optimal solutions.

Investigating self-similarity and heavy-tailed distributions on a large-scale experimental facility

After the seminal work by Taqqu relating self-similarity to heavy-tailed distributions, a number of research articles verified that aggregated Internet traffic time series show self-similarity and that Internet attributes, like Web file sizes and flow lengths, were heavy-tailed. However, the validation of the theoretical prediction relating self-similarity and heavy tails remains unsatisfactorily addressed, being investigated using either numerical or network simulations, or from uncontrolled Web traffic data. Notably, this prediction has never been conclusively verified on real networks using controlled and stationary scenarios, prescribing specific heavy-tailed distributions, and estimating confidence intervals. With this goal in mind, we use the potential and facilities offered by the large-scale, deeply reconfigurable and fully controllable experimental Grid5000 instrument, combined with state-of-the-art estimators, to investigate the predictions observability on real networks. To this end, we organize a large number of controlled traffic circulation sessions on a nationwide real network involving 200 independent hosts. We use a FPGA-based measurement system to collect the corresponding traffic at packet level. We then estimate both the self-similarity exponent of the aggregated time series and the heavy-tail index of flow-size distributions, independently. Not only do our results complement and validate, with a striking accuracy, some conclusions drawn from a series of pioneering studies, but they also bring in new insights on the controversial role of certain components of real networks.

Joint Elastic Cloud and Virtual Network Framework for Application Performance-cost Optimization

Cloud computing infrastructures are providing resources on demand for tackling the needs of large-scale distributed applications. To adapt to the diversity of cloud infrastructures and usage, new operation tools and models are needed. Estimating the amount of resources consumed by each application in particular is a difficult problem, both for end users who aim at minimizing their costs and infrastructure providers who aim at controlling their resources allocation. Furthermore, network provision is generally not controlled on clouds. This paper describes a framework automating cloud resources allocation, deployment and application execution control. It is based on a cost estimation model taking into account both virtual network and nodes managed by the cloud. The flexible provisioning of network resources permits the optimization of applications performance and infrastructure cost reduction. Four resource allocation strategies relying on the expertise that can be captured in workflow-based applications are considered. Results of these strategies are confined virtual infrastructure descriptions that are interpreted by the HIPerNet engine responsible for allocating, reserving and configuring physical resources. The evaluation of this framework was carried out on the Aladdin/Grid’5000 testbed using a real application from the area of medical image analysis.

Fault tolerance for highly available internet services: concepts, approaches, and issues

Fault-tolerant frameworks provide highly available services by means of fault detection and fault recovery mechanisms. These frameworks need to meet different constraints related to the fault model strength, performance, and resource consumption. One of the factors that led to this work is the observation that current fault-tolerant frameworks are not always adapted to existing Internet services. In fact, most of the proposed frameworks are not transport-level- or session-level-aware, although the concerned services range from regular services like HTTP and FTP to more recent Internet services such as multimodal conferencing and voice over IP. In this work we give a comprehensive overview of fault tolerance concepts, approaches, and issues. We show how the redundancy of application servers can be invested to ensure efficient failover of Internet services when the legitimate processing server goes down.

Active Networking Support for the Grid

Grid computing is a promising way to aggregate geographically distant machines and to allow them to work together to solve large problems.After studying Grid network requirements, we observe that the network must take part of the Grid computing session to provide intelligent adaptative transport of Grid data streams.By proposing new intelligent dynamic services, active network can be the perfect companion to easily and efficiently deploy and maintain Grid environments and applications.This paper presents the Active Grid Architecture (A-Grid) which focus on active networks adaptation for supporting Grid environments and applications.We focus the benefit of active networking for the grid on three aspects: High performance and dynamic active services, Active Reliable Multicast, and Active Quality of Service.

Analysis and Experimental Evaluation of Data Plane Virtualization with Xen

Combining end-host, server and router virtualization could offer isolated and malleable virtual networks of different types, owners and protocols, all sharing one physical infrastructure. However, the virtualization of data plane may lead to performance degradation and indeterminism. These arise not only due to additional processing, but also from the sharing of physical resources like memory, CPU and network devices. This paper analyses virtualization from the data plane perspective. We explore the resulting network performance in terms of throughput, packet loss and latency between virtual machines, and also the correspondig CPU cost. The virtual machines act as senders or receivers, or as software routers forwarding traffic between two interfaces in the context of Xen. Our results show that the impact of virtualization on network performance is getting smaller with the successive Xen versions, making this approach a promising solution for data plane virtualization.

Virtualizing and Scheduling Optical Network Infrastructure for Emerging IT Services [Invited]

Emerging IT service providers that aim at delivering supercomputing power available to the masses over the Internet rely on high-performance IT resources interconnected with ultra-high-performance optical networks. To adjust the provisioning of the resources to end-user demand variations, new infrastructure capabilities have to be supported. These capabilities have to take into account the business requirements of telecom networks. This paper proposes a service framework to offer Internet service providers dynamic access to extensible virtual private execution infrastructures, through on-demand and in-advance bandwidth and resource reservation services. This virtual infrastructure service concept is being studied in the CARRIOCAS project and implemented thanks to the scheduling, reconfiguration, and virtualization (SRV) component. This entity handles the service requests, aggregates them, and triggers the provisioning of different types of resources accordingly. We propose to adapt to envisioned heterogeneous needs by multiplexing rigid and flexible requests as well as coarse or fine demands. The goal is to optimize both resource provisioning and utility functions. Considering the options of advanced network bandwidth reservations and allocations, the optimization problem is formulated. The impacts of the malleability factor are studied by simulation to assess the gain.

Optimizing network resource sharing in grids

While grid computing reaches further to geographically separated clusters, data warehouses, and disks, it poses demanding requirements on end-to-end performance guarantee. Its pre-defined destinations and service criteria ease the performance control; however, expensive resources and equipments used by grid applications determine that optimal resource sharing, especially at network access points, is critical. From the resource reservation perspective, this article looks at communication resources shared by grid sites. Two resource request scenarios have been identified, aiming at optimizing the request accept rate and resource utilization. The optimization problems, proven NP-complete, are then solved by heuristic algorithms. Simulation results, aside from showing satisfying results, illustrate the pros and cons of each algorithm

Maximum likelihood estimation of the flow size distribution tail index from sampled packet data

In the context of network traffic analysis, we address the problem of estimating the tail index of flow (or more generally of any group) size distribution from the observation of a sampled population of packets (individuals). We give an exhaustive bibliography of the existing methods and show the relations between them. The main contribution of this work is then to propose a new method to estimate the tail index from sampled data, based on the resolution of the maximum likelihood problem. To assess the performance of our method, we present a full performance evaluation based on numerical simulations, and also on a real traffic trace corresponding to internet traffic recently acquired.

A Scalable Security Model for Enabling Dynamic Virtual Private Execution Infrastructures on the Internet

With the expansion and the convergence of computing and communication, the dynamic provisioning of customized processing and networking infrastructures as well as resource virtualization are appealing concepts and technologies. Therefore, new models and tools are needed to allow users to create, trust and exploit such on-demand virtual infrastructures within wide area distributed environments. This paper proposes to combine network and system virtualization with cryptographic identification and SPKI/HIP principles to help the user communities to build and share their own resource reservoirs. These ideas are implemented in the HIPerNet framework enabling the creation and the management of customized confined execution environments in a large scale context. Based on the example of biomedical applications, the paper focuses on the security model of the HIPerNet system and develops the key aspects of our distributed security approach. Then the paper discusses and illustrates how HIPerNet solutions fulfill the security requirements of applications through different scenarios.