Ahmed Q. Lawey

Distributed Energy Efficient Clouds Over Core Networks

In this paper, we introduce a framework for designing energy efficient cloud computing services over non-bypass IP/WDM core networks. We investigate network related factors including the centralization versus distribution of clouds and the impact of demand, content popularity and access frequency on the clouds placement, and cloud capability factors including the number of servers, switches and routers and amount of storage required in each cloud. We study the optimization of three cloud services: cloud content delivery, storage as a service (StaaS), and virtual machines (VMS) placement for processing applications. First, we develop a mixed integer linear programming (MILP) model to optimize cloud content delivery services. Our results indicate that replicating content into multiple clouds based on content popularity yields 43% total saving in power consumption compared to power un-aware centralized content delivery. Based on the model insights, we develop an energy efficient cloud content delivery heuristic, DEER-CD, with comparable power efficiency to the MILP results. Second, we extend the content delivery model to optimize StaaS applications. The results show that migrating content according to its access frequency yields up to 48% network power savings compared to serving content from a single central location. Third, we optimize the placement of VMs to minimize the total power consumption. Our results show that slicing the VMs into smaller VMs and placing them in proximity to their users saves 25% of the total power compared to a single virtualized cloud scenario. We also develop a heuristic for real time VM placement (DEER-VM) that achieves comparable power savings.

BitTorrent Content Distribution in Optical Networks

In this paper, we extend our previous study on BitTorrent, the most popular peer-to-peer (P2P) protocol, to investigate different aspects related to its energy efficiency in IP over WDM (IP/WDM) networks, validating the power savings previously obtained by modeling and simulation through experimental results. Our contributions can be summarized as follows: First, we compare the energy consumption of our previously proposed energy efficient BitTorrent protocol to that of the original BitTorrent protocol and the client-server (C-S) schemes over bypass IP/WDM networks considering a range of network topologies with different number of nodes and average hop counts. Our results show that for a certain swarm size, the energy efficient BitTorrent protocol achieves higher power savings in networks with lower number of nodes as the opportunity to localize traffic increases. Second, we extend our previously developed energy efficient BitTorrent heuristic enhancing its performance by allowing peers to progressively traverse more hops in the network if the number of peers in the local node is not sufficient. Third, we extend our previously developed mixed integer linear programming model to optimize the location as well as the upload rates of operator controlled seeders (OCS) to mitigate the performance degradation caused by leechers leaving after finishing the downloading operation. Fourth, we compare the power consumption of video on demand (VoD) services delivered using content distribution networks (CDN), P2P, and a promising hybrid CDN-P2P architecture over bypass IP/WDM core networks. A MILP model is developed to carry out the comparison. We investigate two scenarios for the hybrid CDN-P2P architecture: the H-MinNPC model where the model minimizes the IP/WDM network power consumption and the H-MinTPC model where the model minimizes the total power consumption including the network and the CDN datacenters power consumption. Finally, we carry out an experimental evaluation of the original and energy efficient BitTorrent heuristics.

Energy-efficient core networks

In this paper, we optimize core networks to minimize the embodied energy and the operational energy and introduce two measures for energy minimization in core content distribution networks: data compression in optical networks and locality in P2P networks. We investigate the impact of taking the embodied energy of network devices into account in the physical topology optimization and compare it to optimizing the physical topology considering operational energy only. The results show significant embodied energy savings up to 59%, resulting in a total energy saving of 47%. We also investigate energy-efficient data compression for optical networks and how to achieve a trade-off between the power consumption of computational resources and memory required to compress and decompress data and the network power savings. The results show that optimizing the data compression ratios in IP over WDM networks considering the non-bypass approach has saved up to 55% of the networks power consumption. Furthermore, we evaluate the power consumption of BitTorrent, the most popular P2P content distribution protocol, over IP over WDM networks and compare it to client/server (C/S) systems. The results reveal that the power-minimized BitTorrent converges to locality in order to achieve lower power consumption, resulting in 36% power savings compared to the C/S model.

Energy Efficient Tapered Data Networks for Big Data processing in IP/WDM networks

Classically the data produced by Big Data applications is transferred through the access and core networks to be processed in data centers where the resulting data is stored. In this work we investigate improving the energy efficiency of transporting Big Data by processing the data in processing nodes of limited processing and storage capacity along its journey through the core network to the data center. The amount of data transported over the core network will be significantly reduced each time the data is processed therefore we refer to such a network as an Energy Efficient Tapered Data Network. The results of a Mixed Integer linear Programming (MILP), developed to optimize the processing of Big Data in the Energy Efficient Tapered Data Networks, show significant reduction in network power consumption up to 76%.

Renewable energy in distributed energy efficient content delivery clouds

In this paper, we develop a Mixed Integer Linear Programming (MILP) model to study the impact of renewable energy availability, represented by wind farms, on the location of clouds and the content replication schemes of cloud content over IP/WDM networks. In our analysis, we assume that renewable energy is only available to power clouds while the IP/WDM network is powered by non-renewable energy. Our results show that popularity based replication in clouds is the most energy efficient content replication scheme when the clouds are powered only by non-renewable energy sources or when renewable energy availability is limited. With abundant renewable energy, a cloud with a full copy of the content can be built at each node. However, the model should achieve a trade-off between the transmission power losses to deliver renewable energy from wind farms to clouds and the non-renewable power consumption of the IP/WDM network. We discuss this trade-off and show how to optimize the transmission power losses of renewable energy while minimizing the non-renewable network power consumption.

Energy efficient cloud content delivery in core networks

In this paper, we introduce a framework for designing an energy efficient cloud service for content delivery applications over non-bypass IP/WDM core networks. We develop a Mixed Integer Linear Programming (MILP) model to optimize network related factors including the location of the cloud in the network and whether it should be centralized or distributed, and cloud capability factors including the number of servers, switches, routers and amount of storage required at each cloud location. Our results indicate that distributing the cloud into many mini clouds in the network based on content popularity yields 40% total saving in power consumption compared to power un-aware centralized content delivery.

Energy Efficient Resource Provisioning in Disaggregated Data Centres

In this paper we discuss the new paradigm of disaggregated servers (DS) and present our energy efficient heuristic for the energy minimization of virtual machine (VM) placement in data centres implementing the DS approach.

GreenTouch GreenMeter core network power consumption models and results

This paper summarizes the energy efficiency improvement obtained by implementing a number of techniques in the core network investigated by the GreenTouch consortium. These techniques include the use of improved components with lower power consumption, mixed line rates (MLR), energy efficient routing, sleep and physical topology optimization. We consider an example continental network topology, NSFNET, to evaluate the total power consumption of a 2010 network and a 2020 network. The 2020 network results are based on traffic projections, the reductions in the equipment power consumption expected by 2020 and a range of energy saving measures considered by GreenTouch as outlined above. The projections of the 2020 equipment power consumption are based on two scenarios: a business as usual (BAU) scenario and a Green Touch (GT) (i.e. BAU+GT) scenario. The results show that the 2020 BAU scenario improves the network energy efficiency by a factor of 4.8x compared to the 2010 network as a result of the reduction in the network equipment power consumption. Considering the 2020 BAU+GT network where the equipment power consumption is reduced by a factor of 27x compared to the 2010 network, and where sleep, MLR and network topology are jointly optimized, a total improvement in energy efficiency of 64x is obtained.

Energy-efficient peer selection mechanism for BitTorrent content distribution

In this paper, we evaluate the energy consumption of BitTorrent based peer to peer (P2P) content distribution systems in bypass IP/WDM core networks and compare it to client-server (C/S) systems. A Mixed Integer Linear Programming (MILP) model is developed to carry out the comparison. Our results for homogeneous peers with similar upload capacities show that the original BitTorrent, based on random peer selection, has comparable energy consumption to the C/S model. The results also reveal that the power-minimized BitTorrent model achieves 30% energy savings compared to the C/S model as it converges to locality. Furthermore, a heterogeneous BitTorrent system with two upload capacity classes is investigated and the results show a 50% reduction in energy consumption compared to a C/S model. For real-time implementation, we develop a simple heuristic based on the model insights. Comparable power savings are achieved with a reduction of only 13% in the download rate.

Greening big data networks: Volume impact

Tremendous volumes generated by big data applications are starting to overwhelm data centers and networks. Traditional research efforts have determined how to process these vast volumes of data inside datacenters. Nevertheless, slight attention has addressed the increase in power consumption resulting from transferring these gigantic volumes of data from the source to destination (datacenters). An efficient approach to address this challenge is to progressively processing large volumes of data as close to the source as possible and transport the reduced volume of extracted knowledge to the destination. In this article, we examine the impact of processing different big data volumes on network power consumption in a progressive manner from source to datacenters. Accordingly, a noteworthy decrease for data transferred is achieved which results in a generous reduction in network power consumption. We consider different volumes of big data chunks. We introduce a Mixed Integer Linear Programming model (MILP) to optimize the processing locations of these volumes of data and the locations of two datacenters. The results show that serving different big data volumes of uniform distribution yields higher power saving compared to the volumes of chunks with fixed size. Therefore, we obtain an average network power saving of 57%, 48%, and 35% when considering the volumes of 10 - 220 (uniform) Gb, 110 Gb, and 50 Gb per chunk, respectively, compared to the conventional approach where all these chunks are processed inside datacenters only.