Home  
 

Main Menu
Home
About us
Project Description
Quantitative Results
Research Lines
Research Results
Impact on Society
Press room
Contact us
News
Secure Login
Events Calendar
« < October 2017 > »
M T W T F S S
25 26 27 28 29 30 1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31 1 2 3 4 5
Login

Task 3 (UCLM): Equipment of a Grid system with the capability of providing Quality of Service

Leader: Carmen Carrión; Researchers: Mª Blanca Caminero, Agustín Carlos Caminero, Luís Tomás, Dámaso Alcázar

subTask 3.1 Study of the various QoS provision techniques in Grid environments

subTask 3.2 Proposal for improvements in QoS provision techniques in a Grid environment

subTask 3.3 Implementation of a local testbed

subTask 3.4 Implementation of the proposed techniques in the metascheduler

subTask 3.5 Integration of the developed middleware into real systems

1. Brief Description of the Goals

Grid computing enables the aggregation of dispersed heterogeneous resources for supporting large-scale parallel applications in science, engineering and commerce. Current Grid systems are highly variable environments, made of a series of independent organizations sharing their resources, creating what is known as Virtual Organizations (VOs).This variability makes Quality of Service (QoS) highly desirable, though often very difficult to achieve in practice. One reason for this limitation is the lack of control over the network that connects various components of a Grid system. Achieving an end-to-end QoS is often difficult, however, for applications that need a timely response (i.e., collaborative visualization), the Grid must provide users with some kind of assurance about the use of resources – a non-trivial subject when viewed in the context of network QoS.In a VO, entities communicate with each other using an interconnection network – resulting in the network playing an essential role in Grid systems. 

This research line focuses on the provision of QoS in Grid environments by means of efficient meta-scheduling, which considers the network as a key parameter. 

2. Scientific and Technical Developed Activities

Providing Grid systems with schemes to support QoS was the main aim of this research line. To this end, a novel architecture was proposed to provide efficient on-demand and in-advance meta-scheduling mechanisms. Several network-aware heuristic algorithms for meta-scheduling of jobs were developed. One key issue of the proposals was to consider the interconnection network as a first class resource together with the computing resource.

Regarding the meta-scheduling on demand, different techniques were presented to perform meta-scheduling within an administrative domain, and in between domains. Firstly, the buffer-aware meta-scheduler algorithm and the autonomic network-aware meta-scheduler algorithm were proposed for intra-domain on-demand meta-scheduling. The first proposal uses static and dynamic network information such as the number of hops of the network path between a user and a computing resource, and the level of occupation of links of such network path. The second proposal combines concepts from Grid meta-scheduling with autonomic computing, in order to provide users with a more adaptive job management system. Results of these works were published, among others, by A. Caminero et al. in Concurrency and Computation: Practice and Experience, vol 21, issue 13, 2009 and in Advances in Grid Computing pp.49-72, 2011. Moreover, the implementation and tuning of the autonomic algorithm was tested in a real environment and published by L. Tomás et al. in Future Generation Computer Systems, vol. 28, No. 7, 2011.

With respect to multi-domain on-demand meta-scheduling, an extension to the scheduling framework was proposed based on peer-to-peer techniques. More precisely, the proposal is based on Routing Indexes (RI). This way nodes forward queries to neighbors’ that are more likely to have answers. The scalability of the algorithm was demonstrated. The results of the evaluation showed that this proposal outperforms other well-known algorithms in terms of number of queries required for each job, execution time and rate of succeeded jobs. This work, among others, has been published by A. Caminero et al. in Journal of Computer and System Sciences, vol: 77 issue 2, 2011.

Related to meta-scheduling in-advance, a framework was developed on top of Globus and the GridWay meta-scheduler, named SA-Layer. It manages QoS by means of selecting a resource and time period so that QoS requirements of jobs were fulfilled. The system has to be aware of the previous allocation decisions, but no physical reservation of resources is done. Hence, considering the heterogeneity of the Grid system, efficient algorithms based on geometric techniques were implemented. Under this scenario, knowing the status of the system is a must. So, efficient data structures were created to store the needed information and predictive techniques were developed to estimate the future status of the resources (network inclusive). Moreover, several heuristics were implemented to estimate the time needed to complete the job execution in a resource at a specific time in the future. This work has been published by L. Tomás et al. in EuroPar 2010 conference and in Future Generation Computer Systems, vol: 27 issue 5, 2011, among others.        

Additionally, to improve resource utilization by overcoming the fragmentation problem and/or unfavorable previous decisions of the scheduling in-advance, two re-scheduling techniques were proposed. The first one was a reactive technique that moved an already scheduled task in order to try to also accept a new incoming job. The second one was a preventive technique that, from time to time, tried to reduce the fragmentation of the system. Results of these works were published, among others, by L. Tomas et al. in CCGrid 2011 and Grid2011 conferences.

Furthermore, the SA-Layer functionality was improved by the integration of an adaptable fairshare job prioritization system (the FSGrid system, developed by the University of Umeå). Part of this work was published by L. Tomás et al. in Grid 2011 conference.

Let us remark that simulation (based on the GridSim tool) was used to evaluate the performance of the proposals but also a real testbed has been built as a proof-of-concept. As a result, the middleware called SA-Layer is nowadays available as open source to extend the functionality of the well-known GridWay metascheduler.

Finally, it is important to highlight that as a consequence of this work, the Regional Government of Castilla-La Mancha funded the projects PII1C09-0101-9476 and PBI08-0055-2800, being the last one a Project of Technology Transfer to the industry. Moreover, and as scientific results of the research described above Dr. A. Caminero and Dr. L. Tomás got their PhD degree in 2009 and 2012, respectively.


Publications: [Cam09a], [Cam09c], [Cam09d], [Alo09a], [Tom09a], [Cam08a], [Cam08b], [Cam08c], [Cam08d], [Cam08f], [Tom08a], [Tom08c], [Cam07a],[Cam07c],[Cam07d],[Cam07e], [Cam07f], [Cam07g], [Cam07h] 

Projects funded by Public Calls:  [CICYT0609] by  national grants,  [QoSGrid],[MoteGRid] by  regional grants.

External collaborations Academia: Rajkumar Buyya, GRIDSLabOmer Rana, Erik Elmroth

External collaborations Industry: --

Company Agreements: --

PhD dissertations:  Agustín Caminero HerráezLuis Tomás Bolívar

Patents: --