Aristotelis Kretsis

Data Consolidation: A Task Scheduling and Data Migration Technique for Grid Networks

In this work we examine a task scheduling and data migration problem for grid networks, which we refer to as the data consolidation (DC) problem. DC arises when a task needs for its execution two or more pieces of data, possibly scattered throughout the grid network. In such a case, the scheduler and the data manager must select the data replicas to be used and the site where these will accumulate for the task to be executed. The policies for selecting the data replicas and the data consolidating site comprise the data consolidation problem. We propose and experimentally evaluate a number of DC techniques. Our simulation results brace our belief that DC is an important technique for data grids since it can substantially improve task delay, network load and other performance related parameters.

Developing Scheduling Policies in gLite Middleware

We describe our experiences from implementing and integrating a new job scheduling algorithm in the gLite Grid middleware and present experimental results that compare it to the existing gLite scheduling algorithms. It is the first time that gLite scheduling algorithms are put under test and compared with a new algorithm under the same conditions. We describe the problems that were encountered and solved, going from theory and simulations to practice and the actual implementation of our scheduling algorithm. In this work we also describe the steps one needs to follow in order to develop and test a new scheduling algorithm in gLite. We present the methodology followed and the testbed that was set up for the comparisons. Our research sheds light on some of the problems of the existing gLite scheduling algorithms and makes clear the need for the development of new.

Cost and Utilization Optimization of Amazon EC2 Instances

The monitoring and the analysis of public clouds gains momentum, due to their widespread exploitation by individual users, researchers and companies for their daily tasks. We propose an algorithm for optimizing the cost and the utilization of a set of running Amazon EC2 instances by resizing them appropriately. The algorithm, namely Cost and Utilization Optimization (CUO) algorithm, receives information regarding the current set of instances used (their number, type, utilization) and proposes a new set of instances for serving the same load, so as to minimize cost and maximize utilization, or increase performance efficiency. CUO is integrated in Smart cloud Monitoring (SuMo), an open-source tool we develop for collecting monitoring data from Amazon Web Services (AWS) and analyzing them. A number of experiments are performed, using input data that correspond to realist AWS configuration scenarios, which exhibit the benefits of the CUO algorithm.

SuMo: Analysis and Optimization of Amazon EC2 Instances

The analysis and optimization of public clouds gains momentum as an important research topic, due to their widespread exploitation by individual users, researchers and companies for their daily tasks. We identify primitive algorithmic operations that should be part of a cloud analysis and optimization tool, such as resource profiling, performance spike detection and prediction, resource resizing, and others, and we investigate ways the collected monitoring information can be processed towards these purposes. The analyzed information is valuable in driving important virtual resource management decisions. We also present an open-source tool we developed, called SuMo,which contains the necessary functionalities for collecting monitoring data from Amazon Web Services (AWS), analyzing them and providing resource optimization suggestions. SuMo makes easy for anyone to analyze AWS instances behavior, incorporating a set of basic modules that provide profiling and spikef detection functionality. It can also be used as a basis for the development of new such analytic procedures for AWS. SuMo contains a Cost and Utilization Optimization (CUO) mechanism, formulated as an Integer Linear Programming (ILP) problem, for optimizing the cost and the utilization of a set of running Amazon EC2 instances. This CUO mechanism receives information on the currently used set of instances (their number, type, utilization) and proposes a new set of instances for serving the same load that minimizes cost and maximizes utilization and performance efficiency.

Implementing and evaluating scheduling policies in gLite middleware

Grid scheduling algorithms are usually implemented in a simulation environment using tools that hide the complexity of the Grid and assumptions that are not always realistic. In our work, we describe the steps followed, the difficulties encountered and the solutions provided to develop and evaluate a scheduling policy, initially implemented in a simulation environment, in the gLite Grid middleware. Our focus is on a scheduling algorithm that allocates in a fair way the available resources among the requested users or jobs. During the actual implementation of this algorithm in gLite, we observed that the validity of the information used by the scheduler for its decisions affects greatly its performance. To improve the accuracy of this information, we developed an internal feedback mechanism that operates along with the scheduling algorithm. Also, a Grid computation resource cannot be shared concurrently between different users or jobs, making it difficult to provide actual fairness. For this reason we investigated the use of virtualization technology in the gLite middleware. We did a proof‐of‐concept implementation and performed an experimental evaluation of our scheduling algorithm in a small gLite testbed that proves the validity and applicability of our solutions. Copyright

Social-like analysis on Virtual Machine communication traces

We apply social network analysis methods on communication traces, collected from Virtual Machines (VMs) located in computing infrastructures, like a data center. Our aim is to identify important VMs, for example VMs that consume more energy or require more computational capacity, bandwidth, etc, than other VMs. We believe that this approach can handle the large number of VMs present in computing infrastructures and their interactions in the same way social interactions of millions of people are analyzed in todays social networks. We are interested in identifying measures that can locate important VMs or groups of interacting VMs, missed through other usual metrics and also capture the time-dynamicity of their interactions. In our work we use real traces and evaluate the applicability of the considered methods and measures.

Mantis: Cloud-based optical network planning and operation tool

Abstract We present a network planning and operation tool, called Mantis, for designing the next generation optical networks, supporting both flexible and mixed line rate (MLR) WDM networks. Through Mantis, the user is able to define the network topology, current and forecasted traffic matrices, CAPEX/OPEX parameters, set up basic configuration parameters, and use a library of algorithms to plan, operate, or run what-if scenarios for an optical network of interest. Mantis is designed to be deployed either as a cloud service or as a desktop application. Using the cloud infrastructures features Mantis can scale according to the user demands, executing fast and efficiently the scenarios requested. Mantis supports different cloud platforms either public such as Amazon Elastic Compute Cloud (Amazon EC2) and ∼okeanos or private based on OpenStack, while its modular architecture allows other cloud infrastructures to be adopted in the future with minimum effort. The included planning and operation algorithms range from routing and wavelength or spectrum allocation, to equipment (e.g. transponders and regenerators) placement, and CAPEX/OPEX/energy analysis.

An emulation environment for SDN enabled flexible IP/optical networks

Elastic optical networks in conjunction with the Software Defined Networking (SDN) paradigm promise to serve the increasing networking requirements of current and future applications and services, providing on demand and scheduled capacity. We present an emulation environment for SDN-enabled, multi-layer and flexible IP/Optical networks, named Julius, based on Mininet, for the SDN network emulation and LINC-OE, for the optical emulation. Julius also integrates a Path Computation Element (PCE) client that communicates with any standard PCE. Julius can be used for basic and applied research on protocols (e.g., OpenFlow, PCEP) extensions and resource reservation/routing algorithms.

Multi-criteria Virtual Machines Migration Considering the Reconfiguration of Their Logical Topology

We present a methodology, called communication-aware virtual infrastructures (COMAVI), for the concurrent migration of multiple Virtual Machines (VMs) in cloud computing infrastructures, which aims at the optimum use of the available computational and network resources, by capturing the interdependencies between the communicating VMs. This methodology uses multiple criteria for selecting the VMs that will migrate, with different weights assigned to each of them. COMAVI also selects the computing sites/units where the migrating VMs will be hosted, by accounting for the way migration affects the logical (or virtual) topologies formed by the communicating VMs and viewing this selection as a logical topology reconfiguration problem. COMAVI resolves the maximum possible number of VM resource shortages, while tending to minimize the number of migrations performed, the induced network overhead, the logical topology reconfigurations required, and the corresponding service interruptions. We evaluate the proposed method through simulations, where we exhibit their performance benefits.

Mantis: Optical network planning and operation tool

We present a network planning and operation tool, called Mantis, for designing the next generation optical networks, supporting both flex-grid and mixed line rate networks. Through Mantis, the user is able to define the network topology the traffic matrix, the CAPEX/OPEX parameters, setup basic configuration parameters, and use a library of algorithms to plan, operate, or evaluate an optical network of interest. Mantis is designed to be deployed either as a desktop application or as a cloud service. For its execution Mantis can utilize the services of the Amazon Elastic Compute Cloud (Amazon EC2) , however its modular architecture allows other cloud services to be adopted in the future with minimum effort. Using the cloud services Mantis can scale based on the user demands, executing fast and efficiently the scenarios requested. The included planning and operation algorithms range from routing, to equipment placement and to wavelength and spectrum allocation and others.