Enhancing self-scheduling algorithms via synchronization and weighting

Ciorba, Florina M.; Riakiotakis, I.; Andronikos, Theodore; Παπακωνσταντίνου, Γ.; Chronopoulos, Anthony T.

doi:10.1016/j.jpdc.2007.07.003

Cited by 14 publications

(19 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Dynamic scheduling algorithms like CSS [11], GSS [12], TSS [13] adapt their decisions to the state of the system where the state could refer to the number of jobs waiting in the queue to be processed. Most of the scheduling algorithms are implemented using a master-slave or centralized policy where load scheduling is done by a single server.…”

Section: B Load Schedulingmentioning

confidence: 99%

“…Such algorithms are bound to be less reliable than de-centralized algorithms, where load scheduling is done by many servers in the system. In de-centralized algorithms, "synchronization and weighting mechanisms can provide inter-processor communication, thus, enable self-scheduling algorithms to handle efficiently" [16] and improve the performance by 'nested loops with dependencies' [13]. Estimated load information scheduling [17], a de-centralized dynamic load scheduling algorithm estimates system parameters such as "job arrival rate, CPU processing rate, and load on server and balance the load by migrating jobs to buddy server (neighbors) by taking into account the job transfer cost" to perform load scheduling.…”

Section: B Load Schedulingmentioning

confidence: 99%

See 1 more Smart Citation

Load Scheduling in a Cloud Based Massive Video-Storage Environment

Bayyapu¹,

Fischer²

2014

2014 16th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing

View full text Add to dashboard Cite

We propose an architecture for a storage system of surveillance videos. Such systems have to handle massive amounts of incoming video streams and relatively few requests for replay. In such a system load (i.e., write requests) scheduling is essential to guarantee performance. Large-scale data-storage system (LS-DSS) is an emerging hosting facility for video-storage, which has a very high number of writes while most of the videos are never or rarely watched. We discuss the design and implementation of LSDSS and load scheduling in autonomous storage environments called datacenters in LSDSS. A datacenter (DC) is the basic concept in our LSDSS, which has the self-management system to store data efficiently. A LSDSS consists of many DCs organized in a hierarchy fashion, thereby decentralizing load scheduling tasks. Because DC has a simple design, load scheduling is particularly suited for implementation on a real-time video surveillance and allows to make scheduling decisions. We also discuss experimental results that clearly show the advantage of load scheduling over the widely known base load scheduling.2 Referred from web community: accessed June 2014 3

show abstract

Section: B Load Schedulingmentioning

confidence: 99%

Section: B Load Schedulingmentioning

confidence: 99%

Load Scheduling in a Cloud Based Massive Video-Storage Environment

Bayyapu¹,

Fischer²

2014

2014 16th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing

View full text Add to dashboard Cite

show abstract

“…Distributed versions of the loop self-scheduling algorithms (CSS, TSS, FSS, GSS) have been studied before ( [12], [10], [16] and references therein). The approach used in Distributed TSS can be applied to the other self-scheduling algorithms so that they can be efficiently run on non-dedicated heterogeneous systems [10] [16].…”

Section: Determining the Synchronization Frequency For Heterogenmentioning

confidence: 99%

“…The approach used in Distributed TSS can be applied to the other self-scheduling algorithms so that they can be efficiently run on non-dedicated heterogeneous systems [10] [16]. In this work, this is accomplished by transforming the array of m physical heterogeneous processors into an array of A virtual homogeneous processors, each with computational power 1.…”

Section: Determining the Synchronization Frequency For Heterogenmentioning

confidence: 99%

“…. , m. The available computational power is taken as the ratio of the virtual computational power (V P k ) of the worker Unlike the approaches in [10] and [16], here, we use the virtualization method directly in the implementation of the algorithms. In the rest of the paper we will refer to the distributed algorithms applied to arrays of A virtual processors as DCSS, DGSS, DTSS and DFSS.…”

Section: Determining the Synchronization Frequency For Heterogenmentioning

confidence: 99%

See 1 more Smart Citation

Optimal synchronization frequency for dynamic pipelined computations on heterogeneous systems

Ciorba

Riakiotakis

Andronikos

et al. 2007

2007 IEEE International Conference on Cluster Computing

View full text Add to dashboard Cite

Abstract-In this paper we give a theoretical model for determining the synchronization frequency that minimizes the parallel execution time of loops with uniform dependencies dynamically scheduled on heterogeneous systems. Using this model we determine the synchronization frequency that minimizes the estimated parallel time. The accuracy of our method is validated through experiments on a heterogeneous cluster. The results show that the synchronization frequency minimizing the parallel time determined by our method, is very close to the synchronization frequency found experimentally.

show abstract

Towards the optimal synchronization granularity for dynamic scheduling of pipelined computations on heterogeneous computing systems

Riakiotakis

Ciorba

Andronikos

et al. 2012

Concurrency and Computation

View full text Add to dashboard Cite

Loops are the richest source of parallelism in scientific applications. A large number of loop scheduling schemes have therefore been devised for loops with and without data dependencies (modeled as dependence distance vectors) on heterogeneous clusters. The loops with data dependencies require synchronization via cross-node communication. Synchronization requires fine-tuning to overcome the communication overhead and to yield the best possible overall performance. In this paper, a theoretical model is presented to determine the granularity of synchronization that minimizes the parallel execution time of loops with data dependencies when these are parallelized on heterogeneous systems using dynamic self-scheduling algorithms. New formulas are proposed for estimating the total number of scheduling steps when a threshold for the minimum work assigned to a processor is assumed. The proposed model uses these formulas to determine the synchronization granularity that minimizes the estimated parallel execution time. The accuracy of the proposed model is verified and validated via extensive experiments on a heterogeneous computing system. The results show that the theoretically optimal synchronization granularity, as determined by the proposed model, is very close to the experimentally observed optimal synchronization granularity, with no deviation in the best case, and within 38.4% in the worst case.

show abstract

Enhancing self-scheduling algorithms via synchronization and weighting

Cited by 14 publications

References 36 publications

Load Scheduling in a Cloud Based Massive Video-Storage Environment

Load Scheduling in a Cloud Based Massive Video-Storage Environment

Optimal synchronization frequency for dynamic pipelined computations on heterogeneous systems

Towards the optimal synchronization granularity for dynamic scheduling of pipelined computations on heterogeneous computing systems

Contact Info

Product

Resources

About