BOF Discussion: Uploading IEPMBW data to MonALISA - PowerPoint PPT Presentation

1 / 10
About This Presentation
Title:

BOF Discussion: Uploading IEPMBW data to MonALISA

Description:

CAVEATS: Only upload data from relevant and quality probes for example ... ALERTS Info to Upload. Baseline mean and std dev of history' buffer before drop ... – PowerPoint PPT presentation

Number of Views:105
Avg rating:3.0/5.0
Slides: 11
Provided by: conni69
Category:

less

Transcript and Presenter's Notes

Title: BOF Discussion: Uploading IEPMBW data to MonALISA


1
BOF Discussion Uploading IEPM-BW data to MonALISA
  • Connie Logg
  • SLAC
  • Winter 2006 ESCC/Internet2 Joint Techs Workshop
  • February 4-9, 2006

2
Overview
  • Several discussions about uploading IEPM-BW data
    and alerts to MonALISA
  • Consider
  • What parts of data (varies with probe type)
  • Mechanism (every n minutes or as data is taken)
  • What parts of ALERTS?

3
Data Fields
  • Vary with type of probe
  • PING RTT (min, avg, max), Packets sent, Packets
    Received, Packet order (probably not)
  • ? Traceroute list of IP addresses in the
    traceroute

4
Data Fields - Continued
  • IPERF - streams - Window size, throughput
    achieved. With incremental reporting are the
    thruput min, max, avg and std dev of incremental
    values of interest ? (I do not think so)

5
Data Fields - Continued
  • THRULAY TCP total thruput RTT min, max, and
    avg. With incremental reporting are the thruput
    min, max, avg and std dev of incremental values
    of interest? (I do not think so)
  • PATHCHIRP achievable bandwidth min, avg, max of
    chirping steps - NO

6
Data Fields - Continued
  • PATHLOAD min and max of thruput range ?
  • CAVEATS Only upload data from relevant and
    quality probes for example Pathchirp works well
    on some paths, and not so well on others.

7
ALERTS
  • IEPM-BW has an algorithm for auto detection of
    bandwidth drops.
  • Need to choose only alerts on relevant and
    quality probes
  • Do not flood MonALISA with every alert that is
    detected.

8
ALERT Mechanism
  • Baseline data in history buffer has a mean
    and std dev
  • Data less than (mean 2 std dev) is put in
    trigger buffer for examination
  • Trigger buffer also has mean and std dev.
  • When trigger buffer is full and
  • (history_mean trigger_mean)/history_mean meets
    or exceeds the drop threshold, an ALERT is
    generated.

9
ALERTS Info to Upload
  • Baseline mean and std dev of history buffer
    before drop
  • Date and time when drop is detected in trigger
    buffer
  • Percentage of drop from history mean to trigger
    mean
  • Mean and std dev of trigger buffer
  • Values in Trigger buffer (probably not)

10
Guidance/Input Needed
  • What probes, how frequent, and length of history
    and trigger buffers
  • What probe data to upload to MonALISA
  • What alert information to upload to MonALISA
  • How to upload information to MonALISA
  • Other thoughts, comments and suggestions
Write a Comment
User Comments (0)
About PowerShow.com