Check out the new USENIX Web site.
2005 USENIX Annual Technical Conference


Author/Speakers

TRAINING TRACK
Overview | By Day (Sunday, Monday, Tuesday, Wednesday, Thursday) |
All in One File | By Instructor
Wednesday, April 13, 2005
W2 Solaris Kernel Performance, Observability, and Debugging (Day 2 of 2) NEW!
James Mauro and Richard McDougall, Sun Microsystems
9:00 a.m.–5:00 p.m.

See Part 1, T2, for the description of the first day of this tutorial.

Who should attend: System and database administrators, software architects, developers and programmers, performance and systems analysts, and IT architects wanting to obtain a deeper understanding of the key Solaris subsystems, as well as the tools and facilities that can be used to:

  • Observe, trace, and debug to optimize performance
  • Observe, trace, and debug to root-cause aberrent behavior
  • Observe and trace to understand how the application workload interacts with the operating system
  • Better understand the system as a whole
Attendees should have some basic understanding of operating system principles and application performance analysis. Students choosing to attend only Day Two should be familiar with Solaris kernel subsystems and have at least rudimentary knowledge of the bundled tools and utilities and their use.

Applications are becoming more complex every day, and many of the new Solaris features significantly reduce the effort required to administer and anazlyze performance of the entire application and operating system stack.

You may take this class as either a one-day experts class or a two-day complete class. On Day One, we provide an architectual overview of the major Solaris subsystems and an introduction to Solaris performance analysis. On Day Two, we cover advanced topics and spend significant time with hands-on case studies, using the latest tools, including dtrace, mdb, memtool, mdb, trapstat and the Solaris process "ptools."

Topics include:

  • Solaris observability and debugging tools
    • Mastering Solaris DTrace
    • How to debug/monitor with "mdb"
    • Kernel profiling and lock statistics with lockstat
    • Application lock statistics with plockstat
  • Advanced memory architecture and tuning
    • TLB analysis using trapstat
    • Using large pages with the MPSS features
    • NUMA memory allocation and techniques
  • File system performance
    • Tools for measuring and characterizing
    • Analysing file system performance using dtrace
  • Advanced thread scheduling and tools
    • Thread scheduling, parking lots and queues
    • Tracking thread priorities and sleep events
    • Using CPU binding and processor sets
  • Advanced dtrace
    • Attributing network, file I/O to applications
    • Investigating complex inter-process performance problems
    • Tracing unmodified customer applications
  • Workload consolidation and resource management
    • Introduction to tools for workload and resource management
    • Workload measurement
    • Using Solaris resource manager to isolate and control workloads
    • Using Solaris Zones to create Application Containers

James Mauro (T2, W2) is a Senior Staff Engineer in the Performance and AvailabilityJames Mauro Engineering group at Sun Microsystems. Jim's current projects are focused on quantifying and improving enterprise platform availability, including minimizing recovery times for data services and Solaris. Jim co-developed a framework for system availability measurement and benchmarking and is working on implementing this framework within Sun.

Richard McDougall (T2, W2) is a Sun Microsystems Distinguished Engineer who specializes in Richard McDougalloperating systems technology and system performance. He is based at the Menlo Park Performance and Availability Engineering group, where he drives development of performance and behavior enhancements to the Solaris operating system and Sun's hardware architectures. He has led the development of resource management principles, has contributed to the development of virtual memory and file systems within the Solaris operating system, and has architected many tools for analysis, monitoring, and capacity planning. He is the lead author of Resource Management (Prentice Hall). He has written numerous articles and papers on measurement, monitoring, and capacity planning of Solaris systems and frequently speaks at industry and customer technical conferences on the topics of system performance and resource management.

Richard and Jim authored Solaris Internals: Architecture Tips and Techniques (Sun Microsystems Press/Prentice Hall, Feb 2000, ISBN 0-13-022496-0) and are currently collaborating on an update of the book for Solaris 8, as well as volume II.

W3 Implementing LDAP Directories
Gerald Carter, Samba Team/Hewlett-Packard
9:00 a.m.–5:00 p.m.

Who should attend: Both LDAP directory administrators and architects. The focus is on integrating standard network services with LDAP directories. The examples are based on UNIX hosts and the OpenLDAP directory server and will include actual working demonstrations throughout the course.

System administrators today run a variety of directory services, although these are referred to by names such as DNS and NIS. The Lightweight Directory Access Protocol (LDAP) is the up-and-coming successor to the X500 directory and has the promise of allowing administrators to consolidate multiple existing directories into one.

Topics include:

  • Replacing NIS domains
  • Integrating Samba user accounts
  • Integrating MTAs such as Sendmail, Qmail, or Postfix
  • Creating address books for mail clients
  • Managing user access to HTTP and FTP services
  • Integrating with DHCP & DNS servers
  • Scripting with the Net::LDAP Perl module
  • Defining custom attributes and object classes
Gerald Carter (S6, T6, W3) has been a member of the Samba Development Team since 1998. HeGerald Carter has published articles with various Web-based magazines and teaches courses as a consultant for several companies. Currently employed by Hewlett-Packard as a Samba developer, Gerald has written books for SAMS Publishing and is the author of the recent LDAP System Administration for O'Reilly Publishing.

W4 System and Network Monitoring: Tools in Depth

John Sellens, SYONEX
9:00 a.m.–5:00 p.m.

Who should attend: Network and system administrators ready to implement comprehensive monitoring of their systems and networks using the best of the freely available tools. Participants should have an understanding of the fundamentals of networking, familiarity with computing and network components, UNIX system administration experience, and some understanding of UNIX programming and scripting languages.

This tutorial will provide in-depth instruction in the installation and configuration of some of the most popular and effective system and network monitoring tools, including Nagios, Cricket, MRTG, and Orca.

Participants should expect to leave the tutorial with the information needed to immediately implement, extend, and manage popular monitoring tools on their systems and networks.

Topics include, for each of Nagios, Cricket, MRTG, and Orca:

  • Installation—Basic steps, prerequisites, common problems, and solutions
  • Configuration, setup options, and how to manage larger and non-trivial configurations
  • Reporting and notifications—proactive and reactive
  • Special cases—how to deal with interesting problems
  • Extending the tools—how to write scripts or programs to extend the functionality of the basic package
  • Dealing effectively with network boundaries and remote sites
  • Security concerns and access control
  • Ongoing operation
John Sellens (T7, W4, R5) has been involved in system and network administration John Sellens since 1986 and is the author of several related USENIX papers, a number of ;login: articles, and the SAGE Short Topics in System Administration booklet #7, System and Network Administration for Higher Reliability. He holds an M.Math. in computer science from the University of Waterloo and is a chartered accountant. He is the proprietor of SYONEX, a systems and networks consultancy. From 1999 to 2004, he was the General Manager for Certainty Solutions in Toronto. Prior to joining Certainty, John was the Director of Network Engineering at UUNET Canada and was a staff member in computing and information technology at the University of Waterloo for 11 years.

W5 Administering Linux in Production Environments
Æleen Frisch, Exponential Consulting
9:00 a.m.–5:00 p.m.

Who should attend: Both current Linux system administrators and administrators from sites considering converting to Linux or adding Linux systems to their current computing resources. We will be focusing on the administrative issues that arise when Linux systems are deployed to address a variety of real-world tasks and problems arising from both commercial and research-and-development contexts.

Topics include:

  • Recent kernel developments
  • High-performance I/O
    • Advanced filesystems and logical volumes
    • Disk striping
    • Optimizing I/O performance
  • Advanced compute-server environments
    • Beowulf
    • Clustering
    • Parallelization environments/facilities
    • CPU performance optimization
  • High availability Linux: fault tolerance options
  • Enterprise-wide authentication
  • Fixing the security problems you didn't know you had (or, what's good enough for the researcher/hobbyist won't do for you)
  • Automating installations and other mass operations
  • Linux in the office environment

Æleen Frisch (W5) has been a system administrator for over 20 years. She currently looks Aeleen Frischafter a pathologically heterogeneous network of UNIX and Windows systems. She is the author of several books, including Essential System Administration (now in its 3rd edition).

 

?Need help? Use our Contacts page.

Last changed: 31 March 2005 ch