Tactical Capacity Planning for Sysadmins

Full Day
(9:00 am-5:00 pm)

Grand Ballroom D

LISA14: Syseng
S4
Dr. Neil J. Gunther, Performance Dynamics Company
Description: 

Most system administrators are already familiar with diagnosing and tuning computer systems using performance data captured by open source or commercial monitoring tools such as Splunk, Graphite, and BMC Patrol.

This full-day tutorial will show you how to get beyond performance monitoring to performance analysis and capacity planning. These skills are in demand more than ever for sizing (over-engineering can't improve single-threaded performance), procurement (try buying a 10GHz processor), as well as ensuring scalability of large infrastructures used in both private and public clouds.

Since computer hardware has become a mass-produced commodity, its cost no longer drives capacity planning in the strategic sense. The capacity part has become cheap and easy; it's the planning part that requires skill. And capacity planning is not just about the future anymore. Rather, it needs to respond rapidly to the fast-paced changes and tighter budgets of modern business environments. Enter tactical planning: Guerrilla-style capacity planning.

Who should attend: 

Anyone looking for job security by improving their skill set to include capacity management. No specialized background in performance analysis or capacity planning is assumed. A working knowledge of Linux or Unix performance tools will be helpful.

Take back to work: 

The ability to start analyzing performance data you may already have collected to forecast system capacity and predict bottlenecks that can hinder system scalability.

Topics include: 
  • What is performance and capacity management?
  • The Guerrilla approach to capacity planning.
  • Monitoring the volatile technology marketplace for procurement.
  • The three performance metrics you need to know.
  • Who ordered multicores and what are their performance limitations?
  • Statistical forecasting with R.
  • How to establish a capacity line.
  • Queueing analysis for those who can’t wait.
  • How to use PDQ for bottleneck analysis in R, C, Perl, and Python.
  • Quantifying scalability using Amdahl's law and the Universal Scalability Law.
  • Virtualization capacity management from core hyperthreads to cloud hyperservices.
  • Scalability analysis of Xen, VMware and WebLogic virtualized servers.
  • Case studies in capacity planning for large-scale web sites and multi-tier applications.
Presentation Type: 
Training