McFarm Performance Monitoring McPerM

1
McFarm Performance Monitoring(McPerM)

• Richard Thomas
• Benedict College
• Supervisor
• Dr. Hyunwoo Kim
• University of Texas at Arlington

2
Introduction

• My Project
• -Improve McPerM.
• -To re-implement McPerM with updates.
• But before I delve into the specifics of my
  project, lets first get some background
  information

3
Background

• Nature is based on
• four forces gravity, the electromagnetic force,
  the weak
• force, and the strong force.
• Many years of experiment and theory have resulted
  in
• the Standard Model (SM).
• During the 1960s and 1970s, physicists showed
  that
• Electromagnetic and weak forces could be
  unified...
• Strong force and gravity are still separate.

4
The Standard Model

• Years of experimental and theoretical advances
  has lead to the Standard Model.
• However, this is believed to be only an
  approximate theory and current research is being
  done to improve the theory.

5
Fermi National Accelerator Laboratory

• One of the goals of Fermilab is verifying the
  Standard Model.
• To verify it, Fermilab conducts high energy
  experiments.
• This is done with the Tevatron a particle
  accelerator with 2 collision halls, D-Zero and
  CDF.

6
D-Zero Experiment

• The collision events that occur at D-Zero need to
  be compared to Monte Carlo (MC) data.
• MC data are simulated re-creations of
• events that occur. It uses a series of
• mathematical equations to produce sections of
• events and randomize them to ensure their
• authenticity.

7
An Event at D-Zero

• This is an event taken from the D-Zero detector.
• It is also what we are re-creating with the
  Monte-Carlo Simulations.

8
D-Zero Collaboration

• Due to the massive amount of MC data that is
  needed, it would be impossible to create all of
  it at Fermilab.
• D-zero relies on remote institutions to
  supplement MC creation.
• One of the institutions is University of Texas at
  Arlington (UTA) which is a part of the D0
  Southern Analysis Region (D0SAR).

9
McFarm

• UTA has developed a computer software which can
  control MC production that is called McFarm.
• Production itself will be meaningless if it is
  not accompanied by good monitoring software.
• UTA created a number of monitoring software and
  one of which is MCPERM, whose main purpose is to
  show statistical performance of McFarm over a
  period of time.

10
McPerM performs 4 main tasks

• Converts archived information into XML files.
• XML files are collected by main server.
• Files are then converted into McPerM readable
  format.
• Database is parsed and statistics are displayed
  using Ploticus.

11
Components of McPerM

• There are 2 major components to McPerM
• McP_remote This resides on the remote farm.
• McP_server This resides on the main farm which
  hosts the main server.

12
Flow of Data

• This is how data is passed to and from the user.

new_m.htm
Web Front-End
Middle Man
new_cgi_multi.py
Processing Scripts
new_mpdata.py
new_mpspeed_multi.py
new_xml_function.py
Parsing Script
13
Web Front-End
14
Implementation

• Ploticus was used to give graphical
  representations of the users requests.

15
Total Event Generated

• Swift Farm at U. Texas, Arlington

16
DPCC at U. Texas, Arlington
17
CSE farm at U. Texas
18
HEP cluster at Oklahoma University
19
Conclusion

• The purpose of this project was to improve McPerM
  and ensure it maintains steady
• working condition.
• To that extent, the project is complete,
• however further work has to be done to fully
  
• automate McPerM for it to be working
  efficiently.

20
Acknowledgements

• I would like to thank Hyunwoo Kim, Jae Yu, Dr.
  Davenport, Dianne Engram and Elliot McCrory for
  taking the time out to mentor me and making sure
  that my experience here was the best possible.
• I would also like to give a special thank you to
  the SIST Committee.