6
Integrated Real-Time Resource Scheduling Kartik Gopalan and Tzi- cker Chiueh State University of New York Stony Brook {kartik,chiueh}@cs.sunysb Experimental Computer Systems Lab

Integrated Real-Time Resource Scheduling

  • Upload
    liang

  • View
    47

  • Download
    0

Embed Size (px)

DESCRIPTION

Integrated Real-Time Resource Scheduling. Kartik Gopalan and Tzi-cker Chiueh State University of New York Stony Brook {kartik,chiueh}@cs.sunysb.edu. Experimental Computer Systems Lab. Integrated Real-time Resource Scheduling. Motivation for IRS. - PowerPoint PPT Presentation

Citation preview

Page 1: Integrated Real-Time Resource Scheduling

Integrated Real-Time Resource Scheduling

Kartik Gopalan and Tzi-cker Chiueh

State University of New York

Stony Brook

{kartik,chiueh}@cs.sunysb.edu

Experimental Computer Systems Lab

Page 2: Integrated Real-Time Resource Scheduling

Motivation for IRS• Real-time applications

use multiple system resources such as CPU, disk, network and memory.

• Time bound on completion of multiple tasks.

Experimental Computer Systems Lab

Integrated Real-time Resource Scheduling

T

ReadFrame

Transcode

Transmit

Task Precedence Graph

T

Page 3: Integrated Real-Time Resource Scheduling

Challenges• Coordinate allocation

and scheduling of multiple resources.

• Maximize efficiency of system resource usage.

• Key : Per-task delay assignment.

Experimental Computer Systems Lab

Integrated Real-time Resource Scheduling

T

Disk

CPU

Network

Task Precedence Graph

T

T1

T2

T3

Page 4: Integrated Real-Time Resource Scheduling

Contributions

• Optimal delay allocation for Linear TPG

• Iterative relaxation based heuristic for Non-linear TPG.

Experimental Computer Systems Lab

SLt m

j

KjLj

KiLiii

1

*

*

Integrated Real-time Resource Scheduling

Page 5: Integrated Real-Time Resource Scheduling

Proof of Concept Implementation

Experimental Computer Systems Lab

Integrated Real-time Resource Scheduling

Page 6: Integrated Real-Time Resource Scheduling

Work-In-Progress

• Gage : Per-request response time guarantees in Web server cluster.

• Direct solution for non-linear TPG.

Experimental Computer Systems Lab

Integrated Real-time Resource Scheduling