HP-UX performance and tuning

Course overview

This course shows you how to optimize the performance of your computing environment, including multiprocessor and cell-based systems. Extensive hands-on labs allow you to gain experience using standard UNIX and HP-specific tools to monitor, analyze, and tune the performance of HP-UX systems and common network services. The 5-day course is 50 percent lecture and 50 percent hands-on.

Prerequisites

• HP-UX System and Network Administration I (H3064S) and HP-UX System and Network Administration II (H3065S) or for Experienced UNIX System Administrators(H5875S)
• Inside the HP-UX Operating System (H5081S) is advantageous

Audience

Experienced HP-UX system and network administrators

Course objectives

• Explore a methodology for investigating performance issues
• Identify tools used to monitor HP-UX performance
• Identify bottlenecks and potential problems
• Determine appropriate remedial actions to take

Benefits to you

• Effectively utilize the range of performance tools that are available to you
• Learn how to regularly monitor your systems and quickly recognize problems
• Optimize your systems by identifying and removing performance bottlenecks
• Effectively allocate resources such as CPU, memory, disk I/O bandwidth among your critical and lower priority users and applications
• Deliver a guaranteed level of application performance to your end users

Why education services from HP?

• Focus on job-specific skills
• Hands-on practice
• Experienced and best-in-the-field HP instructors
• Comprehensive student materials
• More than 80 training locations worldwide
• Customized on-site delivery

Next Steps

• HP Systems Insight Manager: HP-UX (HB508S)
• HP Capacity Advisor and Global Workload Manager (HF869S)

Detailed course outline

Introduction to Performance

• What is a performance problem
• The "System centric" view of performance
• Measuring performance
• The first rule of interpreting metrics
• Types of performance
• Multiprocessor scaling
• Bottlenecks
• Baseline
• Queuing and response times
• Increasing CPU counts and utilization
• Types of metrics

HP Performance Tools

• HP Performance tools
• Sources of data
• Glance
• GPM - Glance Plus Motif
• Adviser and Alarms
• HP Performance Agent and Manager
• HP Performance Manager
• PRM
• WLM
• gWLM

Unix Performance Tools

• Caliper
• Prospect
• tusc
• lsof
• sar
• top

CPUs and Performance

• Types of CPU bottlenecks
• CPUs and performance
• Data access times
• Tuning for data latency
• Performance and system size
• mpsched
• Launch policies
• Processor sets
• Address translation delays
• Variable page size kernel parameters
• The change attributes command
• Hyperthreading
• Shared caches
• Compiler optimizations

Processes and Performance

• CPU performance problems
• Understanding the "standard" scheduler
• Load balancing
• Managing priorities
• psets
• Priority Inversion
• Interrupt processing
• Looking at CPU Utilization

CPUs Adjusting Performance

• Unix commands to adjust prioritie
• HP Tools and CPU Management
• Process Resource Manager

VxFS I/O Performance

• Layers of I/O
• I/O and performance
• The filesystem layer
• VxFS performance topics
• VxFS inodes and extents
• Defragmenting OnlineJFS filesystems
• Understanding your I/O workload
• Mount options
• Caching controls
• DSYNC
• Concurrent I/O
• vxtunefs
• Performance implications of locks
• Large directories
• Buffered I/O, reading ahead
• Writing behind
• Direct I/O
• Caching

Volume Manager I/O Performance

• Volume managers, introduction
• Mirroring and performance
• Striping
• Multipathing
• Load balancing policies
• Looking at I/O performance in glance
• Examining VxVM performance with vxstat
• How PRM Manages Disk Bandwidth

Disk I/O Performance

• The SCSI layer
• Optimizing I/Os in the SCSI Layer
• Device caching
• Setting low level SCSI parameters
• Immediate reporting with simple disks
• I/O related wait states
• Process system calls

Cell Local Memory and Performance

• LORA
• Uniform Memory access patterns
• Non-Uniform memory access platform
• Locality domains to manage ccNUMA memory
• Configuring CLM
• Impact of I/O locality
• Applications suited to LORA
• LORA with nPars
• LORA with vPars
• LORA with VM
• SAP on LORA
• Java on LORA
• Oracle on LORA
• Strategies for HP NUMA platforms

Memory

• System memory management
• When does memory affect performance
• Virtual memory
• Memory allocation
• vhand, the page daemon
• Memory Resource Groups
• File/Buffer cache paging differences
• Diagnosing memory problems
• Tuning the swap environment
• How PRM manages memory
• Memory File System

Virtual Machine Performance

• Monitoring VMs
• hpvmsar
• HP Integrity VM Manager
• Glance
• Measureware

Java Performance

• Measureware
• Java out of the box
• HPjmeter
• Java Management console
• Recipes for diagnosing problems

Network Performance

• Types of performance
• Latency and response time
• Bandwidth
• Layers within networking