Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.677136
Title: Strategies for the execution of long-term continuous and simultaneous tasks in grids
Author: Haberland, Valeriia
ISNI:       0000 0004 5368 3693
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Increasingly large amounts of computing resources are required to execute resource intensive, continuous and simultaneous tasks. For instance, automated monitoring of temperature within a building is necessary for maintaining comfortable conditions for people, and it has to be continuous and simultaneous for all rooms in the building. Such monitoring may function for months or even years. Continuity means that a task has to produce results in a real-time manner without significant interruptions, while simultaneity means that tasks have to be run at the same time because of data dependencies. Although a Grid environment has a large amount of computational resources, they might be scarce at times due to high demand and resources occasionally may fail. A Grid might be unable or unwilling to commit to providing clients’ tasks with resources for long durations such as years. Therefore, each task will be interrupted sooner or later, and our goal is to reduce the durations and number of interruptions. To find a mutually acceptable compromise, a client and Grid resource allocator (GRA) negotiate over time slots of resource utilisation. Assuming a client is not aware of resource availability changes, it can infer this information from the GRA’s proposals. The resource availability is considered to change near-periodically over time, which can be utilised by a client. We developed a client’s negotiation strategy, which can adapt to the tendencies in resource availability changes using fuzzy control rules. A client might become more generous towards the GRA, if there is a risk of resource exhaustion or the interruption (current or total) is too long. A client may also ask for a shorter task execution, if this execution ends around the maximum resource availability. In addition, a task re-allocation algorithm is introduced for inter-dependent tasks, when one task can donate its resources to another one.
Supervisor: Miles, Simon ; Luck, Michael Mordechai Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.677136  DOI: Not available
Share: