Network Metrics Requirements and Usage for WP1 - PowerPoint PPT Presentation

About This Presentation
Title:

Network Metrics Requirements and Usage for WP1

Description:

Tiziana Ferrari. Network Metrics Requirements and Usage for WP1. 1 ... Tiziana Ferrari. Network Metrics Requirements and Usage for WP1. 4. Closeness i,j(p,th,r) ... – PowerPoint PPT presentation

Number of Views:26
Avg rating:3.0/5.0
Slides: 7
Provided by: Tiziana3
Category:

less

Transcript and Presenter's Notes

Title: Network Metrics Requirements and Usage for WP1


1
Network Metrics Requirements and Usagefor WP1
  • Tiziana Ferrari
  • Tiziana.Ferrari_at_cnaf.infn.it

2
Applicability examples
  • Integration of network info into the scheduling
    policy
  • To define the most suitable Storage Element for
    one or more Computing Elements
  • Triggering of data transfers
  • To gather in a single Storage Element portions of
    data sets partioned among several Ses
  • List of applications to be developed further

3
Network information
  • CloseSE attribute
  • CloseSE (closeness1, SE1), , (closenessn,
    SEn)
  • closeness(CE, SE) max 0ltiltm, 0ltjltn
    closeness(CEi,SEj)
  • Closeness varies in 0,1
  • 0 SE unreachable
  • 1 SE on the same LAN
  • Definitions
  • r Round Trip Time, Rmax max RTT,
  • ploss packet loss frequency
  • th throughput thmax maximum throughput
  • closenessi,j (ploss ,r,th) 0 if ploss ? 0
  • closenessi,j (ploss ,r,th) th / thmax ?
    r/ rmax where 0 lt?
    lt 1 otherwise

4
Closeness i,j(p,th,r)
Fig 1 alpha 1/2
Fig 2 alpha 1/10
5
Triggering of data transfers
  • Problem statement
  • Input data requested by a given job running on CE
    is partioned and portions are stored in different
    SEi
  • Input data needs to be gathered in a single SE
    and the most appropriate one has to be selected
  • Selection creteria
  • Amount of data in each SEi
  • Network distance between two SE
  • Closeness of SEi to CE under consideration
  • SE is the selected Storage Element for acomputing
    element CE iff
  • SE SEi iff D(SEi) minj ? D(SEj) ?
  • where D(SEi) is a SE Distance function.

6
SE Distance function D
  • CASE 1 data in each SEi not known, SEi
    equivalently distant to CE
  • D(SEi) ? 1? j? n, j? i closeness(SEi ,
    SEj )
  • CASE 2 SEi equivalently distant to CE
  • D(SEi) ?max/ ?i ? 1? j? n, j? i
    closeness(SEi , SEj ) ?max/ ?i
  • ?i amount of data in each Sei
  • ?max amount of data requested by a given job
  • CASE 3
  • D(SEi) ? closeness(CE, SEi) ?max/ ?i ? 1?
    j? n, j? i closeness(SEi , SEj )
  • ? constant in 0,1
Write a Comment
User Comments (0)
About PowerShow.com