h4. [Performance Considerations For Developers Utilizing Sun SPARC Enterprise M-Series Servers !Main^download.gif!|http://mapping.sun.com/profile/offer.jsp?id=105]
*by Patrick McGehearty*
{excerpt}November 2008{excerpt}
The Sun SPARC Enterprise® M-Series server product family offers a scalable range of compute power, from rackmount Sun SPARC Enterprise M3000, M4000, and M5000 servers to datacenter Sun SPARC Enterprise M8000 and M9000 servers (Figure 1). These systems implement a common high-performance architecture and utilize the latest SPARC64® processors, helping organizations maximize solution performance.
This Sun BluePrints™ article discusses the characteristics of SPARC64 processors, attributes of each individual Sun SPARC Enterprise M-Series server, and I/O subsystem behavior. Within this document, an emphasis is placed on how system features impact application performance. In addition, the article provides some insights with regard to gaining the best performance from Sun SPARC Enterprise M-Series servers and large computer systems in general. The content of this article assumes a basic understanding of server architecture, multithreaded processors, virtualization technology, and performance tuning principles.
h4. Contents
* Introduction
* Characteristics of SPARC64 VI and SPARC64 VII Processors
** Instruction Compatibility and Extensions
** TLB Structure
** Multithreaded Cores
** Processor to Memory Interface
** Prefetching
* Characteristics of Sun SPARC Enterprise M-Series Servers
** Memory Interconnect
** Memory Latency
** Memory Bandwidth
** SPARC64 VI vs. SPARC64 VII Processor Scalability
** I/O Subsystems
** Migrating to Sun SPARC Enterprise M-Series Servers from Existing SPARC® Platforms
* Solaris OS Improvements for Sun SPARC Enterprise M-Series Servers
** CMT Scheduling
** NUMA Scheduling
** Large Parallel Systems
** Solaris Recommendations
* Advanced Options for Maximizing Performance
** Utilizing the Latest Compilers
** Prefetch Tuning
** Scheduling Optimization
** Network Tuning
* System Tools for Understanding System Behavior
** Obtaining System Configuration Information
** Reporting TLB Misses
** Tracking Lock Contention
** Monitoring System I/O Activity
** Advanced Tools
** Other Tuning Resources
* Summary
* For More Information
* About the Author
* Related Resources
* Ordering Sun Documents
* Accessing Sun Documentation Online
{panel:title=About the Authors|borderStyle=solid|titleBGColor=#F8D583|bgColor=white}
Patrick McGehearty is a Staff Engineer in the Sun Microsystems Performance Technologies Group. During his five years with Sun, he has focused on Solaris OS performance and performance characterization of the Sun SPARC Enterprise M-Series product line with emphasis on how to get the best performance for customer applications. Patrick is currently working on tackling scalability challenges in the Solaris OS in preparation for ever larger systems. Patrick holds a Ph.D. in Computer Science from Carnegie Mellon University. He also holds two patents relating to the interaction of instruction ordering and memory bandwidth optimization.
{panel}
{vote:Rate this blueprint}
Great
Good
Fair
Poor
{vote}
*by Patrick McGehearty*
{excerpt}November 2008{excerpt}
The Sun SPARC Enterprise® M-Series server product family offers a scalable range of compute power, from rackmount Sun SPARC Enterprise M3000, M4000, and M5000 servers to datacenter Sun SPARC Enterprise M8000 and M9000 servers (Figure 1). These systems implement a common high-performance architecture and utilize the latest SPARC64® processors, helping organizations maximize solution performance.
This Sun BluePrints™ article discusses the characteristics of SPARC64 processors, attributes of each individual Sun SPARC Enterprise M-Series server, and I/O subsystem behavior. Within this document, an emphasis is placed on how system features impact application performance. In addition, the article provides some insights with regard to gaining the best performance from Sun SPARC Enterprise M-Series servers and large computer systems in general. The content of this article assumes a basic understanding of server architecture, multithreaded processors, virtualization technology, and performance tuning principles.
h4. Contents
* Introduction
* Characteristics of SPARC64 VI and SPARC64 VII Processors
** Instruction Compatibility and Extensions
** TLB Structure
** Multithreaded Cores
** Processor to Memory Interface
** Prefetching
* Characteristics of Sun SPARC Enterprise M-Series Servers
** Memory Interconnect
** Memory Latency
** Memory Bandwidth
** SPARC64 VI vs. SPARC64 VII Processor Scalability
** I/O Subsystems
** Migrating to Sun SPARC Enterprise M-Series Servers from Existing SPARC® Platforms
* Solaris OS Improvements for Sun SPARC Enterprise M-Series Servers
** CMT Scheduling
** NUMA Scheduling
** Large Parallel Systems
** Solaris Recommendations
* Advanced Options for Maximizing Performance
** Utilizing the Latest Compilers
** Prefetch Tuning
** Scheduling Optimization
** Network Tuning
* System Tools for Understanding System Behavior
** Obtaining System Configuration Information
** Reporting TLB Misses
** Tracking Lock Contention
** Monitoring System I/O Activity
** Advanced Tools
** Other Tuning Resources
* Summary
* For More Information
* About the Author
* Related Resources
* Ordering Sun Documents
* Accessing Sun Documentation Online
{panel:title=About the Authors|borderStyle=solid|titleBGColor=#F8D583|bgColor=white}
Patrick McGehearty is a Staff Engineer in the Sun Microsystems Performance Technologies Group. During his five years with Sun, he has focused on Solaris OS performance and performance characterization of the Sun SPARC Enterprise M-Series product line with emphasis on how to get the best performance for customer applications. Patrick is currently working on tackling scalability challenges in the Solaris OS in preparation for ever larger systems. Patrick holds a Ph.D. in Computer Science from Carnegie Mellon University. He also holds two patents relating to the interaction of instruction ordering and memory bandwidth optimization.
{panel}
{vote:Rate this blueprint}
Great
Good
Fair
Poor
{vote}