SP-XXL
July 13-14, 2006
 
 
 

SP-XXL Workshop

Presenting GPFS

IBM will teach this two-day course on Thursday and Friday, July 13-14, 2006,
in the SCD VisLab at the Mesa Lab
of the National Center for Atmospheric Research, in Boulder, CO.

The course is designed to address the needs of SPXXL sites who use GPFS on IBM AIX and Linux implementations (including Blue Gene). The course content includes an overview of the concepts and constructs of GPFS, the GPFS system environment, tuning, monitoring and problem determination. Using enhanced features of GPFS, including multiclustering capability,
are incorporated in the agenda.    The "Presenting GPFS"  course will be taught if a minimum of 15 attendees confirms participation.   (The maximum number of attendees is 30.)    Exact room information to be provided to attendees.

The course outline includes:
Module 1: Introduction and Overview of GPFS
 - Basic introduction to core concepts, e.g., quorum strategies, raid, data and
                metadata replication, failure groups, disaster recovery
 - An overview of the GPFS cluster topologies, including NSD client/server and RVSD/RSCT
 - Multi-cluster concepts and implementation
 - New Functions / Improvements in Version 3.1, including multi-cluster improvements
                e.g., issues, set up, security, ILM

Module 2: GPFS Environment, Filesystem Constructs, Tuning and Monitoring
 - Building block approach to setting up a GPFS cluster
 - Performance measurements, tuning, tools with focus on mmtrace, mmfsadm, mmpmon
 - Distributed systems administration artifacts, i.e., sys admin in multi-cluster environment
                including configuration issues, migration issues, ID mapping, maintenance,
                security, version 3.1 enhancements in multi-cluster environment

Module 3: Problem Determination
  - Problem determination tools (in-depth)
                - GPFS' logs
                - mmtrace
                - mmfsadm
                - mmpmon
                - use of standard messages (/var/log/messages)
                - gpfs.snap
 - Common mistakes
                - e.g., not setting pagepool, or stat cache size
 - Optimizing data collection to drive quicker problem resolution
                - e.g., minimum data to be collected for opening a problem
 - How to read a GPFS trace

Attendees may bring laptops to use provided internet access in the classroom. "Live" demonstrations are TBD, however numerous examples will be shown to reinforce concepts and recommendations.    

To reserve a seat in this course: Please complete this form by July 7, 2006 and send to:  
Charlotte Lewis (charml@us.ibm.com)

There is no charge for this course. Attendees are responsible for their travel and living expenses. Meals are available on site at the NCAR cafeteria which is open from 7:00 AM to 3:00 PM, serving both breakfast (7:30-9:30 AM) and lunch (11:30 AM-1:30 PM).  


Charlotte Lewis
IBM Marketing, HPC, Public Sector
Office:  877-326-7123
email:  charml@us.ibm.com
www.ibm.com/servers/deepcomputing


Home | Registration | Program | Abstracts | Location | Hotels | Contact