Elastic operations in federated datacenters for performance and cost optimization

Velasco, Luis; Asensio, A.; Berral, Josep Ll.; Bonetto, Edoardo; Musumeci, Francesco; López, Vı́ctor

doi:10.1016/j.comcom.2014.03.004

Cited by 21 publications

(17 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…prime-time [20]. The second traffic profile, named as Datacenter (DC), aggregates traffic of tens of DC to DC connectivity services required for dynamic management activities of distributed DCs, such as DB synchronization or VM migration [21]. The daily traffic pattern of these two profiles is illustrated in Fig.…”

Section: A Metro-flow Traffic Modelling Analysismentioning

confidence: 99%

Dynamic Core VNT Adaptability Based on Predictive Metro-Flow Traffic Models

Morales

Gifre

Paolucci

et al. 2017

J. Opt. Commun. Netw.

Self Cite

View full text Add to dashboard Cite

Abstract-MPLS-over-optical virtual network topologies (VNT) can be adapted to near future traffic matrices based on predictive models that are estimated by applying data analytics on monitored origin-destination (OD) traffic. However, the deployment of independent SDN controllers for core and metro segments can bring large inefficiencies to this core network reconfiguration based on traffic prediction when traffic flows from metro areas are rerouted to different ingress nodes in the core. In such case, OD traffic patterns in the core might severely change thus affecting the quality of the predictive OD models. New traffic models' re-estimation usually entails long time during which no predictive capabilities are available for the network operator. To alleviate this problem, we propose to extend data analytics to metro networks to obtain predictive models for the metro-flows; by knowing how these flows are aggregated into OD pairs in the core, we can also aggregate their predictive models thus accurately predicting OD traffic and therefore, enabling core VNT reconfiguration. To obtain quality metro-flow models, we propose an estimation algorithm that processes monitored data and returns a predictive model. In addition, a Flow Controller is proposed for the control architecture to allow metro and core controllers to exchange metro-flow model information. The proposed model aggregation is evaluated through exhaustive simulation, and eventually experimentally assessed together with the Flow Controller in a testbed connecting premises in CNIT (Pisa, Italy) and UPC (Barcelona, Spain).

show abstract

Section: A Metro-flow Traffic Modelling Analysismentioning

confidence: 99%

Dynamic Core VNT Adaptability Based on Predictive Metro-Flow Traffic Models

Morales

Gifre

Paolucci

et al. 2017

J. Opt. Commun. Netw.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The authors in [17] study the cost savings obtained by inter-connecting the DCs belonging to a federation. In [18], the authors propose an architecture to provide dynamic bandwidth on demand over WDM; in case of failure, dynamic restoration is used.…”

Section: 1mentioning

confidence: 99%

ABNO-driven content distribution in the telecom cloud

Gifre

Tornatore

Contreras

et al. 2017

Optical Switching and Networking

View full text Add to dashboard Cite

Abstract-As current sustained traffic growth is expected to strain capacity of today's metro network, novel content distribution architectures where contents are placed closer to the users are being investigated. In that regard, telecom operators can deploy datacenters (DCs) in metro areas, thus reducing the impact of the traffic going from users to DCs. However, effective solutions to distribute contents to those metro DCs and to synchronize contents among them need to be investigated. In this paper, a hierarchical content distribution architecture for the telecom cloud is investigated: core DCs placed in geographically distributed locations, are interconnected through permanent "per content provider" (CP) virtual network topology (CP-VNT); additionally, metro DCs are placed close to users, and need also to be interconnected with the core DCs. CP's data is replicated in the core DCs through the CP-VNT, while metro-to-core (M2C) anycast connections are established periodically for metro DCs to synchronize contents. Since failures in the interconnection network might disconnect the CP-VNTs, in this paper recovery mechanisms are proposed to reconnect topologies and anycast connections. Topology creation, anycast provisioning and recovery problems are first formally stated and modelled as Integer Linear Programs (ILP). Solving these models, however, becomes impractical for realsized scenarios, so heuristic algorithms are also proposed. Exhaustive simulation results show that the investigated architecture provides significant improvements in both supported traffic and restorability compared to the case of a single core DC. Workflows to implement the algorithms within the Applications-based Network Operations (ABNO) architecture and extensions for PCEP are proposed. Finally, the architecture is experimentally validated in our ABNObased iONE test-bed.Keywords: Datacenter interconnection, Optical networks, Content distribution. INTRODUCTIONCloud-based contents are experiencing an exponential growth in users and data volume. In [1], Cisco forecast a yearly global growth of 23% in datacenter (DC)-to-user traffic, 29% in DC-to-DC traffic, and 22% in the intra-DC traffic. Likewise, Alcatel-Lucent highlights that metro traffic will increase 560% by 2017 mainly driven by IP video and cloud traffic [2], growing two times faster than core network traffic. They conclude that placing contents in DCs closer to users can result in a 41% bandwidth reduction in the metro networks. The resulting distributed DC architecture should come together with the extension of the core network functionalities towards the metro, as proposed in [3].Robust telecom infrastructures are needed to minimize the impact of failures on the services [4]; this is especially relevant in the case of video distribution. In fact, prolonged service outages and disruptions have negatively impacted the confidence of important companies toward cloud environments. Telecom operators are in a vantage position to capitalize on highly-reliable cloud services since they o...

show abstract

“…1) computes periodically the global optimal placement for all the VMs in the federated DCs so as to minimize operational costs whilst ensuring QoE. That centralized architecture has proved to reduce operational costs, when dominated by energy and communication costs [8]. The orchestrator computes VM placement and communicates that to each DC resource manager.…”

Section: 2mentioning

confidence: 99%

Orchestrating connectivity services to support elastic operations in datacenter federations

2015

View full text Add to dashboard Cite

Abstract-Datacenter federations are able to manage appropriately the green energy resources available in each datacenter (DC) thanks to their geographically distributed infrastructure, thus reducing energy expenditure. Scheduling algorithms can compute virtual machine migration, transferring huge amounts of raw data from one DC to another to minimize operational costs and ensuring a certain Quality of Experience (QoE). Because green energy availability greatly depends on weather conditions, in this work we present a statistical model to improve green solar energy availability estimation accuracy and we use it in a mixed integer linear programming (MILP) formulation to compute optimal virtual machine placement. Optical connections can be used to provide connectivity services of enough capacity to support those migrations. In particular, elastic optical networks can provide connections with multi-granular bitrate, which can be adapted on demand. DC resource managers can request optical connections and control their capacity. However, that scheme involves the resource managers to implement algorithms and interfaces to deal with network specifics and complexity. To solve that issue, in this paper we propose coordinating transfer-based inter-DC connectivity services; inter-DC connectivity is requested in terms of volume of data and completion time. We analyze cost savings when each connectivity model is applied in a DC federation. For the sake of a compelling analysis, exhaustive simulation experiments are carried out considering realistic scenarios. Results show that the notificationbased model can save up to 20% of energy costs and more than 40% of communication costs in the evaluated scenarios.Keywords: Federated datacenters, Energy costs minimization, Inter-datacenter networks. INTRODUCTIONThe huge energy consumption of datacenters (DC) requires an elastic resource management, e.g. by turning servers off when they are not used or turning them on to satisfy increments in the demand. Thanks to virtualization, jobs (e.g., web applications) can be encapsulated in virtual machines (VM) and run in the most proper server according to their performance goals. The local resource manager can migrate VMs from one server to another looking for reducing energy consumption while ensuring the committed quality of experience (QoE) [1]. In that regard, the live-migration technique allows migrating VMs from one server to another without stopping them resulting in reduced downtimes. Large Internet companies, such as Google, have their own IT infrastructures consisting in a number of large DCs placed in geographically diverse locations to guarantee the appropriate QoE to users; DCs are interconnected through a wide area network [2]. Using such infrastructure, workloads can be moved among DCs to take advantage of reduced energy cost during off-peak energy periods in some locations while using green energy when it is available in some other locations. Servers are turned off when they are not used, thus minimizing their energy expendit...

show abstract

Elastic operations in federated datacenters for performance and cost optimization

Cited by 21 publications

References 22 publications

Dynamic Core VNT Adaptability Based on Predictive Metro-Flow Traffic Models

Dynamic Core VNT Adaptability Based on Predictive Metro-Flow Traffic Models

ABNO-driven content distribution in the telecom cloud

Orchestrating connectivity services to support elastic operations in datacenter federations

Contact Info

Product

Resources

About