1 / 11

User-Level Performance Monitoring Programme Cees de Laat Hans Blom

Explore the history, projects, and tools of the performance monitoring program at Utrecht University to enhance network efficiency and diagnose issues. The program allows for real-time operation with a focus on network performance improvement.

johnchadwick
Download Presentation

User-Level Performance Monitoring Programme Cees de Laat Hans Blom

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. 1 of 6 User-Level Performance Monitoring Programme Cees de Laat Hans Blom Utrecht University

  2. Contents of this talk 2 of 6 • This space is intentionally left blank

  3. History • DAS project • 4 computer clusters on 4 different universities • Connected via "internet" and ATM • Prove QoS advantage • Management level WWW screen to see if links are up • Dynacore • Remote Control • Video conferencing • Distributed databases • CORBA/Objectivity • Pseudo real time operation • Requirement minimal 10 Mbit/s • Juelich - Utrecht • Weekdays often around 30 - 60 kbit • Was not a local problem, occurred in other German-Dutch projects

  4. Tector 3 of 6

  5. Teleoperation 4 of 6 Jülich Remote Diagnostic TEXTOR Control Visualisation DAQ REMOT Broker Dbs DAQ DAQ Dbs REMOT broker REMOT broker REMOT Broker Dbs A/V A/V A/V Network services

  6. USA line Multi Kingdom Problem 5 of 7 Physics-UU to IPP-FZJ => 7 kingdoms • Netherlands • Physics dept • Campus net • SURFnet • Europe • TEN 155 • Germany • WINS/DFN • Juelich, Campus • Plasma Physics dept 3 ms 2.5 ms 17 ms • Jülich

  7. rTPL package 6a of 7 user account, open for rsh modified netperf + ping Host Host Host idem user account, open for rsh modified netperf + ping idem Host idem Host rsh to initiate measurements collect data WWW server->applet ->brouwser

  8. rTPL @ work 6b of 7 • Every hour: • Controlling host contact host from list • Netperf to other hosts for about 10 sec • 40 ping's to other hosts • Collect data • Contact next host in list • Builds matrix from/to • Not all elements need to be filled • data compressed in zip file • Load on network ~ 1% • User display • User contacts WWW server at controlling host • User gets applet • Applet loads last zipped data file containing a week of data • Older data also loadable • Applet can average per week, per time of day, etc…

  9. Demonstration 6c of 7 This is the tricky part ….

  10. Conclusions 6d of 7 • Toolset rTPL • Flexible package • Unix knowledge essential (scripting) • User friendly • Can pinpoint performance problems if hosts well chosen • Performance • Network performance improved a lot (TEN 1.55 -> TEN 155) • Still not sufficient for full scale virtual lab • Modeling • How does end to end relate to per hop performance tp1 tp2 tp3 Host A Host B Tp = f( tp1 , tp2 , tp3 , ... )

  11. Acknowledgments 7 of 7 • This work is supported by • TERENA Technical Programme • SURFnet bv • European Commission, DG XIII • Telematics Applications Programme Telematics for Research • RE 1008 REMOT • RE 4005 DYNACORE • More info: • http://www.phys.uu.nl/~wwwfi/nettest • http://www.phys.uu.nl/~wwwfi/das • http://www.phys.uu.nl/~wwwfi • http://www.phys.uu.nl/~dynacore QUESTIONS ?

More Related