Uniform Job Scheduling Model for Distributed Processing Environment with Distributed Storage

HEJTMÁNEK, Lukáš and Petr HOLUB. Uniform Job Scheduling Model for Distributed Processing Environment with Distributed Storage. CESNET z.s.p.o., 2004, 13 pp. Technical report 9/2004.

Basic information

Original name

Uniform Job Scheduling Model for Distributed Processing Environment with Distributed Storage

Name (in English)

Uniform Job Scheduling Model for Distributed Processing Environment with Distributed Storage

Authors

HEJTMÁNEK, Lukáš and Petr HOLUB

Edition

13 pp. Technical report 9/2004, 2004

Publisher

CESNET z.s.p.o.

---

Other information

Type of outcome

Účelové publikace

Confidentiality degree

není předmětem státního či obchodního tajemství

References:

URL

Organization unit

Institute of Computer Science

Keywords in English

scheduling; uniform tasks; PO-class scheduling algorithm; network traffic prediction service;

Tags

network traffic prediction service, PO-class scheduling algorithm, scheduling, uniform tasks

---

Abstract

V originále

In this report, we describe components, algorithms and mathematical details for the scheduling jobs with respect to distributed both computing capacity and storage capacity. The work was designed for distributed video processing in Grid environments described in [5]. The video processing features important advantage of having almost exactly uniform job size and thus the scheduling algorithm may belong to PO class under certain assumptions discussed here. Furthermore we provide in-depth discussion of requirements on network traficc prediction services that need to be used to optimize data location with respect to available processing capacity and vice versa.

In English

In this report, we describe components, algorithms and mathematical details for the scheduling jobs with respect to distributed both computing capacity and storage capacity. The work was designed for distributed video processing in Grid environments described in [5]. The video processing features important advantage of having almost exactly uniform job size and thus the scheduling algorithm may belong to PO class under certain assumptions discussed here. Furthermore we provide in-depth discussion of requirements on network traficc prediction services that need to be used to optimize data location with respect to available processing capacity and vice versa.