Platform Performance 27th Edition (Spring 2010)
Platform Performance 27th Edition (Spring 2010)
Chapter 3 (Software Performance) discussed some best practices for building high performance map services, and the importance of selecting the right software technology to support your business needs. Chapter 5 (Performance Fundamentals) provided an overview of capacity planning performance models assuming all hardware platforms were the same. This chapter will focus on hardware platform performance, and share the importance of selecting the right computer technology to support your system performance needs.
- 1 Platform Performance Baseline
- 2 Platform Performance
- 3 ArcGIS Desktop Platform Sizing
- 4 Web Mapping Servers
- 5 Platform Selection Criteria
Platform Performance Baseline
The world we live in today is experiencing the benefits of rapid technology change. Technology advancements are directly impacting GIS user productivity—the way we all deal with information and contribute to our environment. Our ability to manage and take advantage of technology benefits can contribute to our success in business and in our personal life.
To develop a system design, it is necessary to identify user performance needs. User productivity requirements are represented by the workstation platforms selected by users to support computing needs. GIS users have never been satisfied with platform performance, and each year power users are looking for the best computer technology available to support their processing needs. Application and data servers must be upgraded to continue support for increasing user desktop processing requirements.
GIS user performance expectations have changed dramatically over the past 10 years. This change in user productivity is enabled primarily by faster platform performance and lower hardware costs.
Figure 9.1 identifies the hardware desktop platforms selected by GIS users as their performance baseline since the ARC/INFO 7.1.1 release in February 1997. This Performance Baseline History has improved user productivity and expanded acceptance of GIS technology.
Performance Baseline History
Figure 9-2 provides a graphic overview of Intel workstation performance over the past ten years. The chart shows the radical change in relative platform performance since 2000. Technology change introduced by the hardware platform manufacturers represented a major contribution to performance and capacity enhancements over the past 10 years.
The boxes at the bottom of the chart represent the performance baselines selected to support ESRI capacity planning models. These performance baselines were reviewed and updated each year to keep pace with the rapidly changing hardware technology.
The change in hardware performance over the years has introduced unique challenges for customer capacity planning and for software vendors trying to support customer performance and scalability expectations. Understanding how to handle hardware performance differences is critical when addressing capacity planning, performance, and scalability issues.
Figure 9-3 shows how user expectations have changed over the past eight years. An ArcGIS Desktop simple dynamic map display processing time in CY2000 would take almost 6 seconds. That same map display today can be rendered in less than 0.6 seconds - 10 times faster than just 8 years earlier. All of this performance gain can be accounted for by the change in platform technology.
Understanding how to account for platform technology change is fundamental to understanding capacity planning. Figure 9-4 identifies a simple relationship that we have used since 1992 to relate platform performance with capacity planning.
The relationship simply states that if one can determine the amount of work that can be supported by server A (display transactions supported by server A) and identify the relative performance between server A and server B, then one can identify the work that can be supported by server B. This relationship is true for single-core servers (servers with a single computer processing unit) and for multi-core servers with the same number of cores. This relationship is also true when comparing the relative capacity of server A and server B.
Identifying a fair measure of relative platform performance and capacity is very important. Selection of an appropriate performance benchmark and agreement on how the testing will be accomplished and published are all very sensitive hardware marketing issues.
Figure 9-5 shares the mission statement published by the Standard Performance Evaluation Corporation (SPEC), a consortium of hardware vendors established in the late 1980s for the purpose of establishing guidelines for conducting and sharing relative platform performance measures.
The SPEC compute-intensive benchmarks have been used by ESRI as a reference for relative platform capacity metrics since 1992. The system architecture design platform sizing models used in conjunction with these relative performance metrics have supported ESRI customer capacity planning since that time. The SPEC benchmarks were updated in 1996 and 2000 to accommodate technology changes and improve metrics. A new SPEC2006 release was published in 2006 and provides the platform baseline metrics for ESRI system architecture design sizing models starting with the Arc07 2007 baseline (there is normally a 6–12 month overlap in testing and published results once SPEC introduces the new benchmarks). Test results for over 45 vendor platforms were published on the SPEC2000 and the SPEC2006 benchmark sites, and these test results were used to calculate a transfer function between the two benchmark sets for capacity planning purposes. The ratios of the published benchmark results (SPECrate_2000/SPECrate_2006) were plotted on a graph to identify a mean value for the transfer function. Figure 9-6 provides the results of this analysis.
A published SPECrate_int2000 benchmark divided by 2.1 will provide an estimate for the equivalent SPECrate_int2006 benchmark value. A published SPECrate_int2006 benchmark multiplied by 2.1 will provide a translated estimate for the equivalent SPECrate_int2000 benchmark value. It is interesting to notice that the older platform technology has better benchmark values on the SPEC2006 benchmarks than the newer technology (the conversion ratio for the new Intel Xeon 51xx series platforms is over 2.2 while the conversion factor for the older Intel Pentium D platforms are under 2.0).
SPEC provides a separate set of integer and floating point benchmarks. Computer processor core are optimized to support integer or floating point calculations, and performance can be very different between these environments. Testing with the ESRI software since the ArcGIS technology release has followed the integer benchmark results, suggesting the ESRI ArcObjects software predominantly uses integer calculations. The Integer benchmarks should be used for relative platform performance calculations when using ArcGIS software technology.
SPEC also provides two methods for conducting and publishing benchmark results. The SPECint2006 benchmarks measure execution time for a single benchmark instance and use this measure for calculating relative platform performance. The SPECint_rate2006 benchmarks are supported by several concurrent benchmark instances (maximum platform capacity) and measure executable instance cycles over a 24-hour period. The SPECint_rate2006 benchmark results are used for relative platform capacity planning metrics in the ESRI system architecture design sizing models.
There are two results published on the SPEC site for each benchmark, the conservative (baseline) and the aggressive (result) values. The conservative baseline values are generally published first by the vendors, and the aggressive values are published later following additional tuning efforts. Either published benchmark can be used to estimate relative server performance, although the conservative benchmarks would provide the most conservative relative performance estimate (removes tuning sensitivities).
Figure 9-7 provides an overview of the published SPEC2006 benchmark suites. The conservative SPECint_rate2006 benchmark results are used in the ESRI system architecture design documentation as a vendor-published reference for platform performance and capacity planning.
The SPEC performance benchmarks are published on the Web at www.spec.org. The ESRI Capacity Planning Tool (the Capacity Planning Tool will be introduced in Chapter 10) includes a HardwareSPEC workbook that provide a list of published SPECrate_integer benchmarks. The SRint2000 tab includes all vendor published SPECrate_int2000 benchmarks available on the SPEC site. SPEC stopped publishing the SRint2000 benchmarks in January 2007. All the new platform benchmarks are now published on the SPECrate_integer2006 site (SRint2006 tab). The last date the benchmark tab was updated is shown with the link name. A hot link to the SPEC site is included on the top of the Capacity Planning Tool (CPT) hardware tab.
Figure 9-8 identifies the location of the SPEC link on the CPT hardware tab and provides some views of the SPEC site.
Several benchmarks are published on the SPEC Web site. You will need to select and go to the SPECrate2006 Rates and the scroll down to configurable request selection - you can then select specific items that you want included in your display query. I like to include the processor MHz in my display, which was not included in the default selection.
The SRint2006 results tab in the CPT includes an additional column (baseline/core) that I add to the table. This identifies the processing performance of an individual core, a value that is used to estimate relative platform processing performance for a single sequential display. The relative processing performance per core values will be used in comparing user display performance.
Hardware vendor technology has been changing rapidly over the past 10 years. Improved hardware performance has enabled deployment of a broad range of powerful software and continues to improve user productivity. Most business productivity increases experienced over the past 10 years have been promoted by faster computer technology. Technology today is getting fast enough for most user workflows, and faster compute processing is becoming less relevant. Most user displays are generated in less than a second. Access to Web services over great distances is almost as fast. Most of a user's workflow is think time—the time a user spends thinking about the display before requesting more information.
Most future user productivity gains will likely come from more loosely coupled operations, higher capacity network communications, disconnected processing, mobile operations, pre-processed cached maps, and more rapid access and assimilation of distributed information sources. System processing capacity becomes very important. System availability and scalability are most important. The quality of information provided by the technology can make a user's think time more productive.
Hardware processing encountered some technical barriers during 2004 and 2005 which slowed the performance gains experienced between platform releases. There was little user productivity gain by upgrading to the next platform release (which was not much faster), so as a result, computer sales were not growing at the pace experienced in previous years. Hardware vendors searched for ways to change the marketplace and introduced new technology with a focus on more capacity at a lower price. Vendors also focused on promoting mobile technologies, wireless operations, and more seamless access to information. Competition for market share was brutal, and computer manufacturers tightened their belts and their payrolls to stay on top. CY2006 brought some surprises with the growing popularity of the AMD technology and a focus on more capacity for less cost. Intel provided a big surprise with a full suite of new dual-core processors (double the capacity of the single-core chips) while at the same time significant processing performance gains at a reduced platform cost. Hardware vendor packaging (Blade Server technology) and a growing interest in virtual servers (abstracting the processing environment from the hardware) is further reducing the cost of ownership and provide more processing capacity in less space.
Figure 9-9 provides an overview of vendor-published single-core benchmarks for hardware platforms using Intel processor technology.
The Intel Xeon 3200 MHz platform (single-core SPECrate_int2000 = 18 / SPECrate_int2006 = 8.8) was released in 2003 and remained one of the highest-performing workstation platforms available through CY2005. The SPECint_rate2000 benchmark result of 18 was used as the Arc04 and Arc05 performance baseline.
CY2005 was the first year since CY1992 that there was no noticeable platform performance change (most GIS operations were supported by slower platform technology).
There were some noticeable performance gains early in CY2005 with the release of the Intel Xeon 3800 MHz and the AMD 2800 MHz single-core socket processors. An Arc06 performance baseline of 22 (SPECrate_2006 = 10.5) was selected in May 2006. Since May, Intel released the Intel Xeon 5160 4 core (2 chip) 3000 MHz processor, a dual-core chip processor with a single core SPECrate_int2000 benchmark of 30 (SPECrate_int2006 = 13.4) and operating much cooler (less electric consumption) than the earlier 3.8 MHz release. The Arc07 performance baseline of 14 (SPECrate_int2006 = 14) was selected based on the Intel X5160 technology.
Figure 9-10 provides an overview of vendor-published single-core benchmarks for hardware platforms using AMD processor technology.
AMD platforms were very competitive with Intel in the 2004 - 2005 timeframe. Since that time, Intel processor performance improvements have been much more impressive than available AMD alternatives. Intel technology continued to improve in CY2008 and server pricing was even more competitive. Hardware vendors were promoting platforms with dual core chips and reducing the price on lower performance low power configurations. The Xeon 5260 4 core (2 chip) platform (SPECrate_int2006 = 17.5) was selected as the 2008 baseline.
2009 was another great year for performance gains. Intel released a new chip technology that was over 70 percent faster per core than their 2008 release. Hardware vendors stop providing Dual core chip options, and all entry level commodity servers include Quad core or higher capacity chips. The Intel Xeon 5570 8 core (2 chip) 2933 MHz platform was over 3.3 times the capacity of the 2008 baseline at about the same platform cost.
Figure 9-11 provides an overview of vendor-published per core benchmarks for hardware platforms supporting UNIX operating systems.
The UNIX market has focused for many years on large "scale up" technology (expensive high-capacity server environments). These server platforms are designed to support large database environments and critical enterprise business operations. UNIX platforms are traditionally more expensive than the Intel and AMD "commodity" servers, and the operating systems typically provide a more secure and stable compute platform.
IBM (PowerPC technology) is an impressive performance leader in the UNIX environment. Sun has also retained a significant hardware market share with many loyal customers, particularly in the GIS marketplace. Many GIS customers continue to support their critical enterprise geodatabase operations on UNIX platforms. Hardware vendor efforts to reduce cost and provide more purchase options make it important for customers to understand their performance needs and capacity requirements. In the past new hardware included the latest processor technology, and customers would expect new purchases would increase user productivity and improve operations. In today's competitive market place, new platforms do not ensure faster processor core technology. You must understand your performance needs and consider relative hardware performance is selecting the right platform.
Figure 9-12 provides an overview of platform configuration options available on a DELL site. The selected platform is the 2009 recommended optimum ArcGIS Server container machine configuration.
The ideal platform configuration would include the right processor, memory, and hard drive configuration. Configuring a Dell PowerEdge server with two Xeon X5570 quad core 2933 MHz processors, 16GB 1066 MHz memory, 64 Bit standard windows operating system, and dual RAID1 146 GB disk drives cost just over $9,000.
Processors: The X5570 2933 MHz processors provide the best performance/core. Selecting the "Energy Efficient" Xeon X5530 2400 MHz processor reduces the overall cost by just over $1000. What is not shown? Performance of the Xeon X5530 processor is 84 percent of the Intel Xeon X5570 configuration. The reduced cost of the X5504 quad core models attract the budget minded purchaser ($1,450 savings), but the performance is only 55 percent of the X5260 platform. Selecting the right platform to meet your performance and capacity needs is more challenging than ever before. You really need to know your performance need, and the relative performance of platforms is not identified when you make your purchase - you need to know the model number you are looking for, and do your research before you buy, or you may be very disappointed with a platform that is not designed to meet your performance expectations.
Figure 9-13 provides a graphic overview of the current platforms on the market, and shows the relative performance per core for each. The chart also shows the SPEC baseline performance values for reference.
The platforms that run with reduced power are slower than the full power configurations (reduced power means reduced user productivity). Know what you are shopping for before you buy and you will be much happier with the performance of your new platform selection.
ArcGIS Desktop Platform Sizing
Figure 9-14 provides an overview of supported ArcGIS workstation platform technology. This chart shows the Intel platform performance changes experienced over the past five years. The new Xeon W3570 MHz quad-core processor is more than 6 times faster and almost 24 times the capacity of the Pentium D 2400 MHz platform that supported ARC/INFO workstation users in 2004. The advance of GIS technology is enriched by the remarkable contributions provided by ESRI's hardware partners.
Full release and support for Windows 64-bit operating systems provide performance enhancement opportunities for ArcGIS Desktop workstation environments. The increasing size of the operating system executables and the number of concurrent operations supporting GIS operations makes more memory and improved memory access an advantage for ArcGIS Desktop users. Recommended ArcGIS Desktop workstation physical memory with an ArcSDE data source is 2 GB, and 4-6 GB may be required to support large image and/or file-based data sources.
Most GIS users are quite comfortable with the performance provided by current Windows desktop technology. Power users and heavier GIS user workflows will see big performance improvements with the new Xeon i7 or W5570 quad-core technology. Quad-core technology is now the standard for desktop platforms, and although a single process will see little performance gain in a multi-core environment there will be significant user productivity gains by enabling concurrent processing of multiple executables. Parallel processing environments such a 3D image streaming with ArcGIS Explorer 900 and future enhancements with 3D simulation and geoprocessing will leverage the increased capacity of multi-core workstation environments.
Server Platform Sizing Models
Figure 9-15 provides a view of the platform configuration module available with the configuration planning tool (overview presented in Chapter 10). This configuration framework will be used in chapter 11 to share a system architecture design methodology for selecting the right hardware solution to support specific operational performance and scalability needs.
Platform sizing models have been developed and maintained over the years to support ESRI customers with system architecture design planning and proper selection of supported vendor hardware. A fundamental discussion of these models was presented in chapter 7 Performance Fundamentals.
This chapter applies these capacity planning models to current hardware technology and provides some simple engineering charts that can be used for proper hardware selection. The charts presented in this chapter are consistent with the models described in chapter 7, and the sizing charts can be used to validate standard ArcGIS Desktop and Web mapping service platform recommendations generated by the excel based capacity planning tool.
The ESRI software technology patterns have expanded significantly with the ArcGIS 9.3 release, and the charts included in this section address a small range of the fundamental GIS dynamic deployment strategies. The capacity planning tool includes a much broader range of workflow options, and should be used as the primary capacity planning tool for selecting the most optimum technology solutions.
Windows Terminal Server Platform Sizing
Windows Terminal Server supports centralized deployment of ArcGIS Desktop applications for use by remote terminal clients. Figure 9-16 identifies three standard Windows Terminal Server software configurations. A separate platform sizing chart will be provided to address each solution architecture.
Figure 9-17 introduces a standard platform sizing chart that will be used throughout this chapter as a tool to identify peak concurrent users that can be supported with a selected vendor platform configuration. Hardware platforms are represented on sizing charts as a horizontal line. The location of the platform on the chart is determined by the vendor published SPECrate_int2006 benchmark results for the represented platform configuration. The capacity planning models introduced in chapter 5 are represented on these charts.
A platform configuration graphic is included showing the software component installation represented by the sizing chart. The peak displays per minute introduced in chapter 5 are used with the Arc09 component service times to identify specifications required to support each specific GIS workflow. The platform performance specifications are represented by the vendor-published SPECrate_int2006 benchmark on the vertical axis of the sizing chart.
There are three diagonal fans on the sizing chart (WTS Medium 10 DPM, WTS Light 10 DPM, and WTS Light 6 DPM). Each fan includes performance based on three different data sources (Small Shape File, ArcSDE Direct Connect, and ArcSDE Application Server connect). Peak user capacity is determined by dropping down from the intersection of the selected platform configuration (horizontal lines) with the data source configuration on the associated user productivity fan (diagonal lines on the productivity fans).
A WTS Medium workflow is representative of ArcGIS Desktop power users with a medium complexity map display environment (many display layers, heavier display queries, heavy map display complexity). The WTS Light workflow represents higher performance optimized workflow environments.
The concept of user productivity is a parameter introduced with the Arc06 sizing models, and as user workflows generate more complex processing and rich map displays customers may find user productivity will reduce accordingly (i.e., a display generated in 1 second may support a workflow with 10 displays per minute, which a richer display generated in 2 seconds could reduce user productivity to 5 displays per minute). It is important to understand user information display requirements, and provide applications that will support display requirements with optimum use of available system resources (i.e., simple map displays).
Figure 9-18 provides a Capacity Planning Tool Design solution that demonstrates how much hardware technology has changed over the last 5 years and its impact on Windows Terminal Server sizing. The CPT Design is configured with five separate identical ArcGIS Desktop light workflows, each workflow supporting 90 concurrent users hosted on separate Windows Terminal Server platform tier.
2005 platform technology. First user workflow is installed on Tier01 supported by Intel Xeon 2 core (2 chip) 3200 MHz servers. Windows Terminal Server configuration includes 7 platforms, with each platform node supporting up to 14 concurrent ArcGIS Desktop light power users.
2006 platform technology. Second user workflow is installed on Tier02 supported by Intel Xeon 2 core (2 chip) 3800 MHz servers. Windows Terminal Server configuration includes 6 platforms, with each platform node supporting up to 16 concurrent ArcGIS Desktop light power users. Each platform configured with 8 GB physical memory.
2007 platform technology. Third user workflow is installed on Tier03 supported by Xeon 5160 4 core (2 chip) 3000 MHz servers. Windows Terminal Server configuration includes 3 platforms, with each platform node supporting up to 43 concurrent ArcGIS Desktop light power users. Each platform configured with 12 GB physical memory.
2008 platform technology. Fourth user workflow is installed on Tier04 supported by Xeon X5260 4 core (2 chip) 3333 MHz servers. Windows Terminal Server configuration includes 2 platforms, with each platform node supporting up to 54 concurrent ArcGIS Desktop light power users. Each platform configured with 16 GB physical memory.
2009 platform technology. Fifth user workflow is installed on Tier05 supported by Xeon X5570 4 core (1 chip) 2933 MHz servers. Windows Terminal Server configuration includes 1 platform, with each platform node supporting up to 90 concurrent ArcGIS Desktop light power users. Each platform configured with 28 GB physical memory.
Figure 9-19 provides another view of the impact of hardware technology change on Windows Terminal Server sizing. The improvements in processor core performance in conjunction with more processor core per chip have significantly increased server throughput capacity (number of concurrent users supported on a single platform). As the number of concurrent user sessions on a platform increase, the memory requirements must also increase to accommodate the additional concurrent user sessions. Heavier workflows can require more memory per session than lighter workflows. Servers must be configured with sufficient physical memory to take advantage of the higher platform processing capacity.
It is important to take advantage of the Windows 64-bit Operating System for the new Intel platforms, since these higher capacity servers require much more physical memory to handle the high number of current active client sessions. Up to 48 GB of memory is required to take full advantage of the 45-90 concurrent user capacity available with the Xeon X5570 4 core (1 chip) 2933 MHz platforms. 64-bit Operating system improves memory management and provides up to 10 percent performance gains over the Windows 32-bit Server Advanced Operating Systems.
ArcSDE Geodatabase Server Sizing
Figure 9-20 identifies recommended software configuration options for the geodatabase server platforms. The geodatabase transaction models apply to both ArcGIS Desktop and Web mapping service transactions. Normally a geodatabase is deployed on a single database server node, and larger capacity servers are required to support scale-up user requirements. ArcGIS Server 9.2+ includes distributed geodatabase replication services that can be used to distribute instances of a single SDE geodatabase over multiple server nodes.
The ArcSDE and DBMS display processing times (service times) are roughly the same for capacity sizing purposes, so the DBMS Server and ArcSDE Remote Servers platform sizing models are the same. Three platform sizing charts are provided—two are configured for the more common capacity systems (under 2,000 concurrent users) and the other is configured for high-capacity systems (up to 5,000 users concurrent users).
Figure 9-21 shows a platform sizing chart for the ArcSDE Geodatabase Server showing capacity of the more common commodity Windows server platforms.
The geodatabase sizing chart above includes two diagonal fans, one showing performance capacity for light complexity geodatabase configurations that include ArcSDE installed on the DBMS server (can also be used for medium complexity geodatabase direct connect configurations) and the other showing performance capacity for configurations where client applications connect to a light SDE Geodatabase server through a direct connect architecture (ArcSDE executable is not installed on the DBMS server). Database server capacity is doubled when client applications use a geodatabase direct connect architecture.
Geodatabase platform capacity has improved dramatically over the past few years. The Xeon 2 core (2 chip) 3200 MHz platforms introduced in CY2004 would support less than 100 concurrent users in an Application Server Connect (ASC) architecture (SDE on the DBMS platform). Hardware technology performance improvements along with introduction of multi-core processors has increased the peak capacity of Intel Xeon two chip commodity server platforms to over 800 concurrent users (Xeon X5460 8 core (2 chip) 3166(12) MHz platform can support up to 1,800 concurrent users with clients using the recommended Geodatabase Direct Connect architecture.
It is important to take advantage of the Windows 64-bit Operating System for the new Intel platforms, since these higher capacity servers require much more physical memory to handle the high number of current active client connections. Up to 84 GB of memory is required to take full advantage of the 1800 concurrent user capacity available with the Xeon X5570 8 core (2 chip) 2933 MHz platforms.
Figure 9-22 shows an ArcSDE Geodatabase Server platform sizing chart for the smaller capacity UNIX server platforms.
UNIX platforms are having a hard time maintaining their market share within standard GIS enterprise environments. In many cases, platform performance is less than their Intel counterparts and the platform technology cost more. Many critical production database environments continue to be hosted on UNIX platforms, although many IT departments are considering migration to commodity server platforms to reduce cost and improve adaptability.
Figure 9-23 shows a platform sizing chart for the ArcSDE Geodatabase Server demonstrating the high capacity available with the more scalable DBMS platform configurations (capacity up to 5000 concurrent users).
Larger Unix platforms can provide sufficient processing capacity to support well over 5000 concurrent users. The IBM AIX power6 16 core (8 chip) 4700 MHz platform supports up to 3400 concurrent power users - this platform can provide configurations up to 64 core with sufficient processing capacity for over 12,000 concurrent users. For Enterprise GIS environments, sizing servers with thousands of concurrent users is well beyond demonstrated processing loads and network throughput levels the DBMS software and storage infrastructure has experienced for a single server platform environment. A distributed geodatabase architecture provides a lower risk alternative for handling these peak capacity loads.
Figure 9-24 provides another view of the impact of hardware technology change on ArcSDE Geodatabase server sizing. The improvements in processor core performance in conjunction with more processor core per chip have significantly increased server throughput capacity (number of concurrent users supported on a single platform). As the number of concurrent user sessions on a platform increase, the memory requirements must also increase to accommodate the additional concurrent user sessions. Heavier workflows can require more memory per session than lighter workflows. Servers must be configured with sufficient physical memory to take advantage of the higher platform processing capacity.
File Server Platform Sizing
Figure 9-25 provides some general guidelines for File Server Platform Sizing. Query processing is not supported by the File Server platform, thus the platform compute loads are quite light. The primary performance factor determining File Server capacity is network capacity (available bandwidth). Network communication guidelines discussed in Chapter 6 (Network Communications) identify the factors determining file server capacity. It is important to avoid disk contention, and store data across multiple disk using the standard RAID configurations discussed in Chapter 4 (Data Administration). The file server will need to be configured with network cards with adequate bandwidth to accommodate peak traffic flow requirements (network interface controller <NIC> card bandwidth should be at least twice the peak display traffic flow.
ArcGIS Desktop Standard Workflow Performance
Figure 9-26 provides an overview of the display performance targets currently represented in the Standard ESRI Workflows used in the Capacity Planning Tool.
The 14 workflow combinations identified above can be generated from just three Standard ESRI Workflows included on the Capacity Planning Tool workflow tab. The first chart shows workflows using the ArcGIS 9.3.1 Desktop Medium Workstation workflow, while the second chart shows the same workflow with ArcGIS Desktop application supported on a Windows Terminal Server configuration.
Web Mapping Servers
Web mapping services platform sizing guidelines are provided for the ArcIMS and ArcGIS Server software technology. The ArcIMS image service is deployed using the ArcIMS software, and the ArcGIS Server map services are deployed using the ArcGIS Server software. All Web mapping technologies can be deployed in a mixed software environment (they can be deployed on the same server platform together). All mapping services can be configured to access a file data source or a separate ArcSDE database. Geodatabase access can be through direct connect or an ArcSDE server connection.
Web Two-Tier Architecture
Figure 9-27 identifies recommended software configuration options for standard two-tier Web mapping deployments. This configuration option supports the Web server and spatial servers (container machines) on the same platform tier. This configuration is recommended for implementations that can be supported by one- or two-server platforms.
The ArcGIS Server service configurations include performance targets for ArcGIS Server ADF_MXD light and medium dynamic workflows, ArcGIS Server REST MXD and MSD light dynamic services, and ArcIMS Image Service. The ArcGIS Server ADF_MXD light dynamic workflow service times are based on initial performance comparisons between the ArcIMS Image Service and the ArcGIS Server ADF dynamic workflows when generating a simple Image map service. The ArcGIS Server ADF_MXD medium dynamic performance targets represent feedback from customers that are deploying heavier ArcGIS Server map services, where processing loads are twice those required for the simple MXD light dynamic images. Many ArcGIS Server implementations with current technology use the heavier MXD medium dynamic performance targets.
The ArcGIS 9.3 release includes a map service generated with a simple REST API. The ArcGIS Server REST_MXD services do not use the ArcGIS Server Map Editor or Viewer ADF components, and Web processing load is significantly reduced. The ArcGIS Server 9.3.1 REST_MSD light dynamic service uses the new 9.3.1 optimized map document (MSD), using a new graphics rendering engine on the SOC machine. The new graphics rendering engine generates the MSD in roughly the same amount of processing at the legacy ArcIMS image service. The ArcGIS Server REST_MXD and MSD services are included on the performance sizing charts. The ArcIMS Image Server represents our legacy Web mapping software, and is included on the platform sizing charts for reference purposes.
Figure 9-28 provides a two-tier capacity planning chart for ArcIMS and ArcGIS Server platform selection. This sizing chart identifies peak display transaction rates that can be supported on selected Web server platforms (displays per hour and displays per minute are both include on the chart). Peak users are also identified above these values based on user productivity of six displays per minute.
Entry level Xeon X5570 4 core (1 chip) 2933 MHz server platform can support up to 50,000 ArcGIS Server ADF light map displays per hour, 30,000 ArcGIS Server ADF medium map displays per hour, and over 135,000 ArcIMS image map displays per hour. Entry-level ArcGIS Server ADF medium with the Xeon X5570 4 core (2 chip) 2933 MHz platform supports roughly the same peak map transactions per hour that ArcIMS image supported with the Intel Xeon 4 core (2 chip) 3.7 GHz servers just three years ago, ArcGIS Server ADF light supports twice what ArcIMS could do with the same quality display. The new REST MSD light dynamic API extends ArcGIS Server entry level capacity to over 110,000 map displays per hour. Services using pre-cached data sources can support 5-10 times the capacity of the same dynamic services with improved display response times and higher quality maps.
Figure 9-29 provides a CPT Design view of the impact of hardware technology change on ArcGIS Server application server sizing. Service configuration provides REST MXD medium dynamic Web mapping services for peak loads up to 36,000 transactions per hour. The improvements in processor core performance in conjunction with more processor core per chip have significantly increased server throughput capacity (number of peak service transactions supported on a single platform).
2005 platform technology. Tier01 is a single tier ArcGIS Server map service configuration with a local FGDB data source, supported by Intel Xeon 2 core (2 chip) 3200 MHz servers. ArcGIS Server configuration includes 7 platforms, with each platform node supporting up to 7,125 ArcGIS Server map service transactions per hour.
2006 platform technology. Tier02 is a single tier ArcGIS Server map service configuration with a local FGDB data source, supported by Intel Xeon 2 core (2 chip) 3800 MHz servers. ArcGIS Server configuration includes 6 platforms, with each platform node supporting up to 8,240 ArcGIS Server map service transactions per hour. Each platform configured with 4 GB physical memory.
2007 platform technology. Tier03 is a single tier ArcGIS Server map service configuration with a local FGDB data source, supported by Xeon 5160 4 core (2 chip) 3000 MHz servers. ArcGIS Server configuration includes 3 platforms, with each platform node supporting up to 21,661 ArcGIS Server map service transactions per hour. Each platform configured with 8 GB physical memory.
2008 platform technology. Tier04 is a single tier ArcGIS Server map service configuration with a local FGDB data source, supported by Xeon X5260 4 core (2 chip) 3333 MHz servers. ArcGIS Server configuration includes 2 platforms, with each platform node supporting up to 28,276 ArcGIS Server map service transactions per hour. Each platform configured with 16 GB physical memory.
2009 platform technology. Tier05 is a single tier ArcGIS Server map service configuration with a local FGDB data source, supported by Xeon X5570 4 core (1 chip) 2933 MHz servers. ArcGIS Server configuration includes 2 platforms, with each platform node supporting up to 46,790 ArcGIS Server map service transactions per hour. Each platform configured with 8 GB physical memory.
Web Mapping Performance Changes
Web mapping services have experienced dramatic performance changes over the past 4 years. These performance enhancements improve Web user productivity and reduce deployment cost. Some of these performance changes were due to expanding software deployment options and others were due to improved hardware processing speed and platform capacity changes.
Figure 9-30 provides an overview of available 2005 - 2006 technology. ArcIMS was the primary Web mapping choice, with new ArcGIS Server Web mapping applications providing options for deploying much richer Web mapping applications. Display performance ranged from 1 - 3 seconds over remote 1.5 Mbps connections. Typical entry level ArcIMS Image Service configurations supported peak throughput of 8,00 to 18,000 transactions per hour, while richer ArcGIS Server map services supported about half this capacity.
Figure 9-31 provides an overview of available 2007 - 2008 technology. ArcGIS Server Web mapping applications were gaining market share, with ArcIMS mapping services retaining a major market share. ArcIMS and ArcGIS Server ADF display performance improved slightly ranging from 1 - 2.5 seconds over remote 1.5 Mbps connections. Typical entry level ArcIMS Image Service configurations supported peak throughput of 21,000 to 55,000 transactions per hour, while richer ArcGIS Server ADF applications supported about half this capacity.
ArcGIS Server REST services and a new Map Cache data source were introduced in 2008 expanding ArcGIS Server development options. ArcGIS Server REST services improved entry level Web dynamic mapping services throughput capacity by over 20 percent over similar ArcGIS Server ADF deployments. Map Cache data source reduced dynamic server loads to almost zero (preprocessed map services), with remote client display performance determined primarily be network bandwidth. Map cache tiles would be retained in the local browser cache, providing very fast Web mapping experience for clients working in an established local area.
Figure 9-32 provides an overview of available 2009 technology. This was the first year ArcGIS Server providing a dynamic Web mapping deployment pattern that outperformed the ArcIMS Image service, removing any remaining ArcIMS benefits over ArcGIS Server and providing a broad range of proven functional benefits encouraging ArcIMS migration to current Web mapping software technology. Web mapping performance improved to a range from 0.5 - 2.0 seconds over remote 1.5 Mbps connections. Typical entry level ArcIMS Image Service configurations supported peak throughput up to 90,000 transactions per hour, while similar ArcGIS Server dynamic mapping applications supporting peak throughput loads up to 118,000 transactions per hour.
ArcGIS Server REST MSD services and improved Map Cache base layer mashups were introduced in 2009 enhancing and expanding ArcGIS Server development options. ArcGIS Server REST MSD services improved entry level Web dynamic mapping services throughput capacity by over 100 percent, significantly enhancing performance and quality of maps provided by ArcGIS Server REST MXD deployments. Map Cache base layer mashups significantly reduced dynamic map layer transaction loads, introducing a new back-office data management strategy (pre-processing map cache basemap layers) for publishing fast interactive mapping services.
The 10 workflow combinations above provide a representative subset of the Standard ESRI Workflows for Web Mapping Services included on the Capacity Planning Tool workflow tab. The CPT Calculator can generate hundreds of customer workflow performance targets based on software technology selection and map service configuration parameters. The ArcGIS Server 9.3.1 software technology options, along with platform performance improvements of over 70 percent per core, make 2009 a record breaking year for Web service performance improvements.
Network bandwidth is currently one of the primary factors impacting Web client display performance. Server processing load variations of the different ArcGIS Server deployment patterns have a secondary impact on client display performance. Server platform technology (processor performance and platform capacity) along with the software technology and display performance parameters determine platform sizing and peak server throughput capacity.
Platform Selection Criteria
Figure 9-33 provides a summary of the factors contributing to proper hardware selection. These factors include the following:
Platform Performance: Platform must be configured properly to support user performance requirements. Proper platform technology selection based on user performance needs and peak system processing loads significantly reduces implementation risk. ESRI performance sizing models establish a solid foundation for proper hardware platform selection. The ESRI Capacity Planning Tool automates the System Architecture Design analysis, providing a framework for coupling enterprise GIS user requirements analysis with system architecture design and proper platform technology selection.
Purchase Price: Cost of hardware will vary depending on the vendor selection and platform configuration. Capacity Planning Tools can identify specific technology required to satisfy peak system processing needs. Pricing should be based on the evaluation of hardware platforms with equal display performance platform workflow capacity.
System Supportability: Customers must evaluate system supportability based on vendor claims and previous experience with supporting vendor technology.
Vendor Relationships: Relationships with the hardware vendor may be an important consideration when supporting complex system deployments.
Total Life Cycle Costs: Total cost of the system may depend on many factors including existing customer administration of similar hardware environments, hardware reliability, and maintainability. Customers must assess these factors based on previous experience with the vendor technology and evaluation of vendor total cost of ownership claims.
Establishing specific hardware technology specifications for evaluation during hardware source selection significantly improves the quality of the hardware selection process. Proper system architecture design and hardware selection provide a basis for successful system deployment.