Invited Talks

Invited Talks

by

Neil J. Gunther
Performance Dynamics Consulting

This list is a push-down stack with the most recent talks at the Top.

YEAR 2006
March 30, 2006
Guerrilla Capacity Planning
Keep the Joint Running, Las Vegas, Nevada
March 7, 2006
Quantum Design Rules for Optical Engineers
Santa Clara Valley, California
IEEE Lasers and Electro-Optics Society
February 7, 2006
Scalability on a Stick
Northern California CMG
Sponsored by SAS Institute, San Francisco, California

YEAR 2005
April 1, 2005
The Millennium Performance Problems
Connecticut CMG, Cromwell, Connecticut
March 31, 2005
The Millennium Performance Problems
Greater Boston CMG, Framingham, Massachusetts

YEAR 2004
April 5-17, 2004
Benchmarking Blunders
Workshop on Performance and Reliability II, Menlo Park, California

YEAR 2003
November 5, 2003
How to Extract Performance Information Nuggets from a Mire of Performance Measurements
RadView Webinar Series, Burlingon, Massachusetts

YEAR 2002
October 23, 2002
Celebrity Boxing (and Sizing): Alan Greenspan vs. Gene Amdahl
PERNET Seminar, San Francisco State University, California
September 4, 2002
Linux Load Average Not Your Average Average
Linux Users of Victoria, Telstra Auditorium, Level 1, 242 Exhibition St, Melbourne, Australia.
April 15, 2002
The Entomology of Server Sizing Models
Sun Microsystems, Menlo Park, California

YEAR 2001
July 3, 2001
The Dynamics of Performance Transients in Large-scale Communication Networks
Telematica Instituut, Enschede, HOLLAND
June 26, 2001
Techniques for Website Sizing and Scaling
Computer Science Dept., University of Geneva, Geneva, SWITZERLAND
April 19, 2001
Techniques for Website Sizing and Scalability
BayLISA, Incyte Genomics, Palo Alto, CALIFORNIA

YEAR 2000
July 24, 2000
UNIX Fair-share Schedulers: What You Don't Know Can Hurt You
Computer Science Building (26), Monash University, Melbourne, AUSTRALIA
The latest generation of commercial UNIXes (e.g., AIX, HPUX, and Solaris) offer facilities for automatically managing the consumption of system resources such as processors, disks, and memory. This is accomplished by re-engineering the generic time-share scheduler to behave as a so-called "fair-share" scheduler. This simple administrative "set it and forget it" approach to resource control belies the potential adverse impact on performance as measured by user response time and service level targets. None of the commercial implementations discuss these issues nor do they provide tools that are suitable for avoiding them. In this talk I will review the concept of fair-share scheduling and the motivations behind deploying it. Then, I will present several share allocation scenarios based on the PDQ modeling tool together with the surprising performance consequences. In closing, I will say something about the future of resource management in UNIX.
July 11, 2000
Next Generation Linux: What Not to Do!
Linux User Group of Victoria, Melbourne, AUSTRALIA



File translated from TEX by TTH, version 3.38.
On 15 May 2006, 19:56.