Thursday, 14 December 2017

Session 3: Monitoring and Execution Support

Towards a Monitoring Feedback Loop for Cloud Applications

Authors:

Piotr Bar (Imperial College London)
Rudy Benfredj (Imperial College London)
Jonathon Marks (Imperial College London)
Deyan Ulevinov (Imperial College London)
Bartosz Wozniak (Imperial College London)
Giuliano Casale (Imperial College London)
William J. Knottenbelt (Imperial College London)

Abstract:

Performance monitoring is fundamental to track cloud application health and service-level agreement compliance, but with the emergence of multi-cloud deployments, it may become increasingly important also to create a feedback loop between runtime operation in multi-clouds and design-time reasoning. This is because the developer needs to acquire more information on the specific performance features of a cloud platform to better leverage its specificities.

To support this goal, we have developed a set of open source components that extract quality-of-service (QoS) data from a target Java application using JMX, aggregate it in a time-series database, and finally deliver it in a prototype Java dashboard that may be integrated in a development environment, such as Eclipse, to display either live or historical QoS data. The architecture is not only limited to collection, aggregation, and display of QoS data, but it also allows the evaluation of hierarchical queries expressed using the Performance Trees graphical language. It is our intention that this will provide a cloud-independent uniform interface for developers to specify monitoring queries. Initial evaluation suggests that Cube on MongoDB provides appropriate scalability for this application.

DOI: 10.1145/2462326.2462336

Full text: PDF

[#][]

Automatic Virtual Machine Clustering based on Bhattacharyya Distance for Multi-Cloud Systems

Authors:

Claudia Canali (University of Modena and Reggio Emilia)
Riccardo Lancellotti (University of Modena and Reggio Emilia)

Abstract:

Size and complexity of modern data centers pose scalability issues for the resource monitoring system supporting management operations, such as server consolidation. When we pass from cloud to multi-cloud systems, scalability issues are exacerbated by the need to manage geographically distributed data centers and exchange monitored data across them. While existing solutions typically consider every Virtual Machine (VM) as a black box with independent characteristics, we claim that scalability issues in multi-cloud systems could be addressed by clustering together VMs that show similar behaviors in terms of resource usage. In this paper, we propose an automated methodology to cluster VMs starting from the usage of multiple resources, assuming no knowledge of the services executed on them. This innovative methodology exploits the Bhattacharyya distance to measure the similarity of the probability distributions of VM resources usage, and automatically selects the most relevant resources to consider for the clustering process. The methodology is evaluated through a set of experiments with data from a cloud provider. We show that our proposal achieves high and stable performance in terms of automatic VM clustering. Moreover, we estimate the reduction in the amount of data collected to support system management in the considered scenario, thus showing how the proposed methodology may reduce the monitoring requirements in multi-cloud systems.

DOI: 10.1145/2462326.2462337

Full text: PDF

[#][]

Managing Elasticity Across Multiple Cloud Providers

Authors:

Fawaz Paraiso (University of Lille & Inria Lille - Nord Europe)
Philippe Merle (University of Lille & Inria Lille - Nord Europe)
Lionel Seinturier (University of Lille & Inria Lille - Nord Europe)

Abstract:

In the context of cloud computing, elasticity is the capacity to scale computing resources up and down easily. Currently, most Platforms as a Service (PaaS) manage application elasticity within a single cloud provider. However, the not so infrequent issue of cloud outages has become a concern that hinders the availability of cloud-based applications. The most promising solutions to this issue are those based on the federation of multiple clouds. In this paper, we present a Multi-Cloud-PaaS architecture. We show how this architecture can be used for managing elasticity across multiple cloud providers.

DOI: 10.1145/2462326.2462338

Full text: PDF

[#][]

A Broker-based Framework for Multi-Cloud Workflows

Authors:

Foued Jrad (Karlsruhe Institute of Technology)
Jie Tao (Karlsruhe Institute of Technology)
Achim Streit (Karlsruhe Institute of Technology)

Abstract:

Computational science workflows have been successfully run on traditional HPC systems like clusters and Grids for many years. Today, users are interested to execute their workflow applications in the Cloud to exploit the economic and technical benefits of this new emerging technology. The deployment and management of workflows over the current existing heterogeneous and not yet interoperable Cloud providers, however, is still a challenging task for the workflow developers. In this paper, we present a broker-based framework for running workflows in a multi-Cloud environment. The framework allows an automatic selection of the target Clouds, a uniform access to the Clouds, and workflow data management with respect to user Service Level Agreement (SLA) requirements. Following a simulation approach, we evaluated the framework with a real scientific workflow application in different deployment scenarios. The results show that our framework offers benefits to users by executing workflows with the expected performance and service quality at lowest cost.

DOI: 10.1145/2462326.2462339

Full text: PDF

[#][]