Luis Carlos Erpen De Bona

A Survey on Cloud Computing Elasticity

Elasticity is a key feature in the cloud computing context, and perhaps what distinguishes this computing paradigm of the other ones, such as cluster and grid computing. Considering the importance of elasticity in cloud computing context, the objective of this paper is to present a comprehensive study about the elasticity mechanisms available today. Initially, we propose a classification for elasticity mechanisms, based on the main features found in the analysed commercial and academic solutions. In a second moment, diverse related works are reviewed in order to define the state of the art of elasticity in clouds. We also discuss some of the challenges and open issues associated with the use of elasticity features in cloud computing.

AutoElastic: Automatic Resource Elasticity for High Performance Applications in the Cloud

Elasticity is undoubtedly one of the most striking characteristics of cloud computing. Especially in the area of high performance computing (HPC), elasticity can be used to execute irregular and CPU-intensive applications. However, the on- the-fly increase/decrease in resources is more widespread in Web systems, which have their own IaaS-level load balancer. Considering the HPC area, current approaches usually focus on batch jobs or assumptions such as previous knowledge of application phases, source code rewriting or the stop-reconfigure-and-go approach for elasticity. In this context, this article presents AutoElastic, a PaaS-level elasticity model for HPC in the cloud. Its differential approach consists of providing elasticity for high performance applications without user intervention or source code modification. The scientific contributions of AutoElastic are twofold: (i) an Aging-based approach to resource allocation and deallocation actions to avoid unnecessary virtual machine (VM) reconfigurations (thrashing) and (ii) asynchronism in creating and terminating VMs in such a way that the application does not need to wait for completing these procedures. The prototype evaluation using OpenNebula middleware showed performance gains of up to 26 percent in the execution time of an application with the AutoElastic manager. Moreover, we obtained low intrusiveness for AutoElastic when reconfigurations do not occur.

An Analysis of Public Clouds Elasticity in the Execution of Scientific Applications: a Survey

Elasticity can be seen as the ability of a system to increase or decrease the computing resources allocated in a dynamic and on demand way. It is an important feature provided by cloud computing, that has been widely used in web applications and is also gaining attention in the scientific community. Considering the possibilities of using elasticity in this context, a question arises: “Are the available public cloud solutions suitable to provide elasticity to scientific applications?” To answer the question, in a first moment we present a survey on the use of cloud computing in scientific scenarios, providing an overview of the subject. Next, we describe the elasticity mechanisms offered by major public cloud providers and analyzes the limitations of the solutions in providing elasticity for scientific applications. As the main contribution of the article, we also present an analysis over some initiatives that are being developed to overcome the current challenges. In our opinion, current computational clouds are developing rapidly but have not yet reached the necessary maturity level to meet all scientific applications elasticity requirements. We expect that in the coming years the efforts being taken by numerous researchers in this area identify and address these challenges and lead to better and more mature technologies that will improve cloud computing practices.

A Survey on SDN Programming Languages: Toward a Taxonomy

Network devices have always been considered as configurable black boxes until the emergence of software-defined networking (SDN). SDN enables the networks to be programmed according to the user requirements; furthermore, it allows the network to be easily modified to suit transient demands. However, how do we program the network? SDN-compliant switches offer a low-level interface that makes programming error-prone. The controllers provide application programming interfaces, but still low-level, limited, and inflexible. High-level languages have the potential to be a better alternative to program the network. There exist several SDN programming languages implementing different sets of functionalities, and focusing on solving various issues. In face of all this diversity, no published paper outlines a pragmatic view allowing to compare the SDN languages. This paper presents a systematic survey of up-to-date OpenFlow-based SDN programming languages. Our approach relies on a taxonomy comprising all prominent features found in those languages. A detailed review of the existing works is carried out investigating the foundational parts of the languages with their contributions. Examples are discussed to illustrate the fundamental abstractions. Last, all gathered information is summarized, discussing the main ongoing research efforts and challenges. Future abstractions and features to be incorporated into the next generations of SDN programming languages are also considered.

A SNMP-Based Virtual Machines Management Interface

This paper presents Virtual-Machines-MIB, a MIB (Management Information Base) directed to virtual machines management through SNMP (Simple Network Management Protocol). Virtual-Machines-MIB aims to define a standard interface of virtual machines management, allowing the management of several virtual machines monitors, like Xen, KVM and VMWare, with a common SNMP management tool. Different from previous virtual machines management MIBs, which allows the manager to perform only monitoring operations, Virtual-Machines-MIB allows to perform control operations, like create, delete, restart, turn on, pause and shut down virtual machines. It is also possible to use the proposed solution to change a virtual machines name, amount of RAM, virtual CPUs and virtual storage drives. Practical results are presented using ordinary SNMP management tools performing KVM and Xen management. To do this, SNMP agents which support Virtual-Machines-MIB were developed and installed on KVM and Xen hosts. These SNMP agents are based on NET-SNMP public domains agent, that was extended to support Virtual-Machines-MIB using libvirt API.

On Modelling Virtual Machine Consolidation to Pseudo-Boolean Constraints

Cloud Computing is a new paradigm of distributed computing that offers virtualized resources and services over the Internet. To offer Infrastructure-as-a-Service (IaaS) many Cloud providers uses a large data center which usage ranges 5% to 10% of capacity in average. In order to improve Cloud data center management and resources usage a Virtual Machine (VM) consolidation technique can be applied to increase workloads and save energy. Using VM consolidation, we introduce an artificial intelligence consolidation based in Pseudo-Boolean (PB) Constraints to find a optimal consolidation. To evaluate our PB consolidation approach we used the DInf-UFPR and Google Cluster scenario and the formulas are solved with two state-of-the-art solvers.

Constructing Elastic Scientific Applications Using Elasticity Primitives

Elasticity can be seen as the ability of a system to increase or decrease the computing resources allocated in a dynamic and on demand way. Considering its importance, some mechanisms to explore elasticity have been proposed by public providers and by academy. However, these solutions are inappropriate to provide elasticity for scientific applications or are limited to a specific programming model. In this context, we present Cloudine, a platform for development of elastic scientific applications based in simple elasticity primitives. These primitives enable the dynamic allocation and deallocation of resources in several levels, ranging from nodes of a virtual cluster, to virtual processors and memory of a node. Using this basic building blocks it is possible to develop applications in different models. The Cloudine effectiveness is demonstrated in the experiments, where two elastic applications in different models were developed.

HyperBone: A Scalable Overlay Network Based on a Virtual Hypercube

This paper presents HyperBone, an overlay network based on a virtual hypercube that offers services such as monitoring and routing, allowing the execution of distributed applications across the Internet hypercubes are scalable by definition, presenting several properties such as symmetry and logarithmic diameter, that are advantageous for distributed and parallel applications. HyperBone nodes run the distributed virtual hypercube algorithm (DiVHA) in order to maintain the topology. DiVHA keeps the hypercube properties even when the number of nodes is not a power of two, or under a dynamic fault situation, in which nodes fail and recover continuously, leaving and joining the system. HyperBone was implemented and experimental results are presented, obtained from the execution of a set of MPI parallel applications on a virtual hypercube spread across the world built with PlanetLab nodes.

Finding stable cliques of PlanetLab nodes

Users of large scale network testbeds often execute experiments that require a set of nodes that behave and communicate among themselves in a reasonably stable pattern. In this work we call such a set of nodes a stable clique, and introduce a monitoring strategy that allows their detection in PlanetLab, a non-trivial task for such a large scale dynamic network. Nodes monitor each other by sampling the RTT (Round-Trip-Time) and computing its variation. Based on this data and a threshold, pairs of nodes are classified as stable or unstable. A set of graphs is generated, on which maximum sized cliques are computed. Three experiments were conducted in which hundreds of nodes were monitored for several days. Results show the unexpected behavior of some nodes, and the size of the maximum stable clique for different time windows and different thresholds.

A programming-level approach for elasticizing parallel scientific applications

We present an approach for exploring cloud elasticity in scientific applications.Elasticity control is embedded in the application source code.Elastic applications can adapt their execution environment according to demands.This approach creates a new paradigm for design and development of applications.We present a framework to support the construction of elastic applications. Elasticity is considered one of the fundamental properties of cloud computing. Several mechanisms to provide the feature are offered by public cloud providers and in some academic works. We argue these solutions are inefficient in providing elasticity for scientific applications, since they cannot consider the internal structure and behavior of applications. In this paper we present an approach for exploring the elasticity in scientific applications, in which the elasticity control is embedded in application source code and constructed using elasticity primitives. This approach enables the application itself to request or to release its own resources, taking into account the execution flow and runtime requirements. To support the construction of elastic applications using the presented approach, we developed the Cloudine framework. Cloudine provides all components necessary to construct and execute elastic scientific applications. The Cloudine effectiveness is demonstrated in the experiments where the platform is successfully used to include new features to existing applications, to extend functionalities of other elasticity frameworks and to add elasticity support to parallel programming libraries.