Windows HPC Server 2008 Fundamentals for the IT Professional

Prerequisites:

Course Description:

This course provides attendees with the knowledge and skills necessary to setup and maintain a high-performance cluster running Microsoft Windows HPC Server 2008. Intended for the Windows IT Professional, our goal is to quickly bring you up to speed on the field of high-performance computing, and then dive into the details of Windows HPC Server 2008. Topics include architectural overview, job scheduling, policies and configuration, HPC application types, cluster setup, networking topologies, data storage options, performance evaluation, and maintenance issues.

The course follows a lecture/lab format, providing every attendee with hands-on experience setting up a cluster, submitting jobs, evaluating performance, configuring policies and templates, installing software, and performing routine maintenance tasks.

This course includes the following modules:

Introduction to High-Performance Computing and HPC Server 2008

• Motivation for HPC
• Brief product history of CCS and HPCS
• Brief overview of HPC Server 2008 components, job submission, scheduler
• HPC resources: cores vs. sockets vs. nodes
• Product differentiators
• The goal of your developers linear speedup

The HPCS Job Scheduler in Depth

• Throughput vs. performance
• Jobs vs. Tasks
• Job and task states
• The definition of a failed job / task, auto re-starting upon failure
• Default scheduling policies
• Job-level vs. task-level policies
• The impact of job priorities and job preemption
• Job resources and dynamic growing / shrinking

The Main Components of an HPC Server Cluster

• Headnode vs. compute nodes vs. broker nodes
• Windows Server 2008 + HPC Pack
• SQL Server
• Active directory and credentials management
• Network topologies
• Data storage options: SMB, NAS, PFS, ...
• Job Scheduler and the job queue

Interfacing with HPC Server From Command Prompts to Calling the HPC Server API

• Cluster Manager
• Job Manager
• Job Description Files
• clusrun
• Console window
• PowerShell
• Scripts
• Programmatically via HPCS API

Understanding how Clients Will Use the Cluster

• What applications can run on the cluster
• What software technologies will developers typically use
• Sequential apps
• Parametric sweep
• SOA applications
• Multi-threaded apps
• MPI apps
• UNIX apps

Configuring HPC Server

• Node groups
• Job templates
• Job preemption
• Dynamic resource allocation
• Submission and activation filters
• Job history, job restarting
• Applying configurations to HPC users and groups

Basic Cluster Setup, from Hardware to Software

• Hardware, physical and virtual
• Software: Windows Server 2008 editions, HPC Pack, HPC Pack SDK
• Active Directory integration
• Common groups, local directories, and network shares
• Network topologies, DNS, and DHCP
• Runtimes, software, and tools commonly needed by developers
• Supporting remote debugging and tracing

Advanced Setup and Integration Options

• Windows Deployment Services
• High-speed networking hardware and drivers, e.g. InfiniBand
• Configuring headnode failover with Windows Server Failover Clustering
• Enabling support for Open Grid Forums basic web profile
• Cluster integration (with other HPCS clusters, and Linux clusters)

Cluster Troubleshooting and Performance Tuning

• Cluster Manager diagnostics
• Common job and task failures from the command-line to licensing
• App-specific failures with MPI and SOA
• Security issues
• Remote desktop as a debugging technique
• Windows HPC Server application event log
• SOA tracing and performance tuning
• Tuning tools: perfmon, xperf, MPI Ping-Pong, and The Lizard

Maintenance of your HPC Server 2008 Cluster

• Cluster Manager charts and reports
• Monitoring via System Center Operations Manager
• Performance maintenance: MPI Ping-Pong, The Lizard, and uSane
• Synchronizing software patches to the cluster
• Headnode maintenance
• SQL Server maintenance
• Compute and broker node maintenance
• Windows Update
• Automating maintenance runs