Adaptive Cloud Technologies

Towards true governance in hybrid cloud

We aim to create improved understanding and technologies for organisations to work together effectively using integrated, scalable and adaptable software. This complexity is further exacerbated by the emerging cloud computing model, where traditional enterprise workloads may need to be 'bursted' over the cloud for cost and agility reasons. Our team is examining the processes, adaptation techniques and software systems being used within financial institutions, government agencies, and networks of organisations across industries. The research and technology outcomes of our work can also be transferred to other domains such as health and e-science.

The features

• Extensible monitoring engine covering local server and cloud, at the infrastructure, platform and application levels.
• Novel monitoring in hybrid cloud: for data consistency, elasticity and cost.
• Diagnose and suggest optimal system configurations
• Performance modeling and resource optimisation
• Auto generation of system reconfiguration workflows.
• Migration and disaster recovery in the public cloud.

The Benefits

• Cloud computing allows for renting 3rd party's computing resources and cut operational cost.
• Challenges for enterprises such as lack of visibility and control covering both in-house and cloud.
• We offer suites for monitoring, diagnosing and system adaptation.

The Big Picture

Components under development

An Extensible Monitoring Engine for Hybrid Cloud

Deploying applications across heterogeneous environments and platforms requires efficient monitoring solutions to keep track of their behavior and performance. Monitoring hybrid cloud is challenging as the monitoring activities cross over various technologies and platforms from the application level to the IT infrastructure level. In this project, we built an extensible monitoring engine for hybrid cloud environments allowing to leverage the right monitoring tools for resources involved in hybrid cloud applications, whether they are on-premises or in a public/private cloud.

The monitoring engine consists of the following main components:

• The Monitoring Tools correspond to a set of monitoring tools available to monitor a set of resource in-house and on the cloud.
• The Monitoring Adapters are developed to adapt monitoring tools and agents to a common interface and to convert their metric data into a common data model managed by the monitoring engine. The proposed architecture is plugin-based such that additional monitoring adapters are pluggable to the monitoring engine for the support of additional monitoring tools and agents. A plugin API has been developed and offers a quick and easy development process of additional adapters using Java. Additionally, every monitoring adapter provides the necessary information to enrich a Knowledge Base of monitoring tools. Such information consists of a kind of "declaration" of all the resource types it supports, the metrics associated to every resource, as well as their associated meta-information such as the meaning (description), the thresholds to consider for alerts, the impact of a metric on the state of a resource, and the mapping of each metric to a common ontology. Therefore, a monitoring adapter is not only converting metric data from one format to another, it also provide rich semantic about every monitoring tool and its output.
• The Monitoring Engine is in charge of executing the monitoring adapters at regular interval to collect the converted metric data and store it into the Data Store. It also collect the meta-information provided by monitoring adapters and store them into the Knowledge Base. Finally, the monitoring engine initiates periodically data analysis routines provided by the Analytics Engine described later in this section.
• The Data Store is a key-value database storing efficiently metric values provided by the monitoring adapters. A key-value entry in the Data Store corresponds to a resource-metric pair where the resource is represented using a unique resource identifier and the metric value is a numerical value provided by the monitoring adapter.
• The Knowledge Base is a relational database storing meta-information provided by the monitoring adapters. The Knowledge base data model is quite simple but yet powerful to represent entities involved in the monitoring process including tools, resources, metrics, thresholds, etc. Note that the Knowledge Base is not in charge of storing metric data. That is undertaken by the Data Store which is a key-value store, efficient to store huge amount of monitoring data.
• The Dashboard is the key component to enable user interaction with the monitoring engine and it allows to display data collected from the monitoring adapters and results generated by the analytics engine. The dashboard is the access point and it consists of one single interface presenting a uniform information and allowing the analysis of monitoring data collected from various tools and platforms.

Related Projects

• (Current) Continuous Availability for Cloud Based Applications
• (Current) CloudDB AutoAdmin
• (Past) Business Adaptation and Interoperation (e.g., the Business-IT Adaptation technology can be used for the Decision Support Engine)

People

Activity

At the ATP Research Lab