logo EDITE Dallal BELABED
Identité
Dallal BELABED
État académique
Thèse soutenue le 2015-04-24
Sujet: Informatique en nuages et son impact sur les futurs protocoles du réseau Internet
Direction de thèse:
Laboratoire:
Voisinage
Ellipse bleue: doctorant, ellipse jaune: docteur, rectangle vert: permanent, rectangle jaune: HDR. Trait vert: encadrant de thèse, trait bleu: directeur de thèse, pointillé: jury d'évaluation à mi-parcours ou jury de thèse.
Productions scientifiques
oai:hal.archives-ouvertes.fr:hal-01120825
On traffic fairness in data center fabrics
International audience
—A present challenge in data center networks (DCN) is to better understand the impact of novel flattened and modular DCN architectures on congestion control protocols, and vice-versa. One of the major concerns in congestion control being the fairness in the offered throughput, the impact of the additional path diversity and forwarding features, brought by the novel DCN architectures and protocols, on the throughput of individual endpoints (servers) and aggregation points (edge switches) is unclear. This paper attempts to answer these open questions. Specifically, how best is the allocation of the competing elastic demand flows and how is this allocation impacted with the increase in capacity? We provide an optimization formulation of the problem based on the proportional fairness principle of TCP. We conducted a series of test scenarios on the fat-tree data center topology by considering load balancing and link capacities for different network cases in order to present our analysis on the results. We observed that the traffic allocation fairness is primarily impacted by the weights associated with the TCP implementation in use.
IEEE International Conference on Cloud Networking (CloudNet), https://hal.archives-ouvertes.fr/hal-01120825 IEEE International Conference on Cloud Networking (CloudNet), , Oct 2014, Luxembourg, Luxembourg. <10.1109/CloudNet.2014.6968966>Conference papers 2014-10-08
oai:hal.archives-ouvertes.fr:hal-01120823
Impact of virtual bridging on virtual machine placement in data center networking
International audience
—The increasing adoption of virtualization techniques has recently favored the emergence of useful switching functions at the hypervisor level, commonly referred to as virtual bridging. In the context of data center network (DCN) consolidations, for VMs colocated in the same virtualization server, virtual bridging becomes very useful to offload inter-VM traffic from access and aggregation switches, at the expense of an additional computing load. DCN consolidations typically chase traffic engineering (TE) and energy efficiency (EE) objectives, and both should be affected by virtual bridging, but it is not intuitive to assert whether virtual bridging acts positively or negatively with respect to TE and EE that should also depend on the DCN topology and forwarding techniques. In this paper, we bring additional understanding about the impact of virtual bridging on DCN consolidations. First, we present a repeated matching heuristic for the generic multi-objective DCN optimization problem, with possible multipath and virtual bridging capabilities, accounting for both TE and EE objectives. Second, we assess the impact of virtual bridging on TE and EE in DCN consolidations. Extensive simulations show us that enabling virtual bridging has a negative impact when EE is the goal and multipath forwarding is adopted, while it leads to important gains, halving the maximum link utilization, when TE is the DCN consolidation goal. I. INTRODUCTION The recent achievement of x86 virtualization by advanced software techniques allows attaining virtualization of server and network functions at competitive performance-cost trade-offs with respect to legacy solutions. The increasing adoption of virtualization techniques has recently favored the emergence of useful switching functions at the hypervisor level, commonly referred to as virtual bridging. In the context of data center network (DCN) consolidations, for VMs colocated in the same virtualization server, virtual bridging becomes very useful to offload inter-VM traffic from access and aggregation switches, at the expense of an additional computing load on the physical server. DCN consolidations typically chase traffic engineering (TE) and energy efficiency (EE) objectives, and both should be affected by virtual bridging, but it is not intuitive to assert whether virtual bridging acts positively or negatively with respect to TE and EE that should also depend on the DCN topology and forwarding techniques. In this paper, we bring additional understanding about the impact of virtual bridging on DCN consolidations. The literature on DCN consolidation problems is extensive. Often referred to as DCN optimization, VM placement or virtual network embedding, the various propositions at the state of the art often have a narrow scope, with a set of constraints so that it is not possible to jointly adopt EE
26th International Teletraffic Congress (ITC) https://hal.archives-ouvertes.fr/hal-01120823 26th International Teletraffic Congress (ITC), Sep 2014, Karlskrona, Sweden. France. <10.1109/ITC.2014.6932945>Conference papers 2014-09-09
oai:hal.archives-ouvertes.fr:hal-01120818
Impact of Ethernet Multipath Routing on Data Center Network Consolidations
International audience
—With the advent of network virtualization, data center networking is reaching a high level of management complexity. Indeed, interconnection networks in data center networks (DCN) are no longer just based on flat over-provisioned pipes, but are increasingly facing traffic engineering (TE) issues that commonly characterize long-haul provider networks. TE objectives, however, are opposite to energy efficiency (EE) objectives commonly chased by virtual machine (VM) consolidations. Moreover, the specific topologies of DCNs and the systematic use of multipath forwarding make the joint TE and VM consolidation optimization complex. The contribution of this paper is twofold. First, we propose a repeated matching heuristic for the DCN optimization problem with multipath capabilities, which also scales well for large topologies without discarding both TE and EE objectives. Second, we assess the impact of multipath forwarding on TE and EE goals. Extensive simulations show us that multipath forwarding is beneficial only when EE is not the primary goal in network-aware VM consolidations, and that it can be counterproductive when instead the EE is the primary goal of such optimizations.
The 4th international workshop on Data Center Performance (DCPerf), The 34th International Conference on Distributed Computing Systems (ICDCS https://hal.archives-ouvertes.fr/hal-01120818 The 4th international workshop on Data Center Performance (DCPerf), The 34th International Conference on Distributed Computing Systems (ICDCS, Jun 2014, Madrid, Spain. <10.1109/ICDCSW.2014.10>Conference papers 2014-06-30
Soutenance
Thèse: Optimisation des réseaux virtuels : conception et évaluation
Soutenance: 2015-04-24
Rapporteurs: Michele NOGUEIRA    Maier GUIDO