System Implementation 31st Edition
Fall 2012 System Implementation 31st Edition
Successful system implementation requires good leadership and careful planning. A good understanding of every component of the system is critical in putting together an implementation strategy. Enterprise IT environments involve integration of a variety of vendor technologies. Interoperability standards within commercial software environments are voluntary, and even the most simple system upgrade must be validated at each step of the integration process.
Enterprise GIS environments include a broad spectrum of technology integration. Most environments today include a variety of hardware vendor technologies including database servers, storage area networks, Windows Terminal Servers, Web servers, map servers, and desktop clients,—all connected by a broad range of local area networks, wide area networks, and Internet communications. All these technologies must function together properly to support a balanced computing environment.
A host of software vendor technologies including database management systems, ArcGIS Desktop and ArcGIS Server software, Web services, and hardware operating systems—all integrated with existing legacy applications. Data (including business layers, basemap layers, and Imagery) and user applications are added to the integrated infrastructure environment to support the final implementation. The result is a very large mixed bag of technology that must work together properly and efficiently to support user workflow requirements.
The integration and implementation of distributed computer technology have become easier over the years as interface standards have matured. At the same time, enterprise environments have become larger and more complex. The complexity and risk associated with an enterprise system deployment are directly related to the variety of vendor components required to support the final integrated solution.
Centralized computing solutions with a single database environment are the easiest environments to implement and support. Distributed computer systems with multiple distributed database environments can be very complex and difficult to deploy and support. Many organizations are consolidating their data resources and application processing environments to reduce implementation risk and improve administrative support for enterprise business environments.
- 1 GIS Staffing
- 2 System Architecture Deployment Strategy
- 3 System Testing
- 4 Systems Integration Management
- 5 Managing Technology Change
- 6 Conclusion
- 7 Previous Editions
People are the most valuable asset to any organization. GIS managers have the special opportunity and challenge to bring people across their organization together to share responsibility for building and maintaining enterprise GIS operations. In most cases, building a GIS is not a single department responsibility – and in many cases the quality of the GIS depends on people outside the organization. GIS brings people together because they depend on each other in establishing the framework for building geographic information products that support their business needs.
An enterprise GIS will not be successful without senior management support and leadership. The first step in establishing your GIS team is to understand the mission of the organization, the mandate and role of GIS within the organization, and to establish executive sponsorship for the planning effort. Figure 11-1 identifies the key membership for the senior Management Committee.
The Management Committee is established to formally manage project scope, schedule and budget, and availability of planning resources. They will be kept informed of progress during the planning effort and will be responsible for directing the GIS implementation phase that follows.
GIS Planning Team
Building a good GIS starts with good planning. Figure 11-2 identifies the management structure for the GIS planning team. The GIS planning team must include business representatives that understand the organization, the business processes (how work gets done), and the business information needs. The team also needs technical experts that understand GIS technology and the systems required to make them work.
A small organization will need a GIS Manager, a Business Representative to represent each major business process, and a GIS technical expert responsible for developing the proper technical architecture. Larger enterprise organizations will need to establish this basic structure (GIS Manager <Site Facilitator>, Business Representatives, and GIS Technical Expert) at each site location with a Geographic Information Officer (GIO) or enterprise level GIS Manager coordinating the overall planning effort. A GIS planning expert or consultant will be responsible for the overall planning process and will report to the senior GIS Manager/GIO. During the GIS planning phase, all of the planning team members should formally report to the GIS Manager for the duration of the planning.
GIS Implementation Team
Once GIS planning is complete, the focus will transfer to implementing the plan. Formal implementation approval and funding will come from the Management Team. Figure 11-3 identifies the management structure for the GIS implementation team.
The GIS Manager, GIS technical expert, and business representatives will normally transition to the implementation team. An additional GIS data specialist and a senior software engineer will be included on the team to manage the GIS implementation tasks. For large enterprise implementations, each GIS Site Facilitator will have a complete implementation team for their specific site implementation. During implementation, the Site Facilitators will report to line management staff working under direction of the Management Committee. The senior GIS Manager/GIO will manage GIS operations across the enterprise coordinating with the various site facilitors.
GIS Organizational Structure
Figure 11-4 shows an overview of a traditional GIS matrix organization. Enterprise GIS operations must be supported by an executive committee with influence and power to make financial and policy decisions for the GIS user community. The GIS manager should establish a coordinating committee responsible for providing technical direction and leadership. Committee leaders should chair working groups assigned and aligned with each technical discipline to address organizational issues and report on system status. The user community should be represented throughout the review process.
A formal organizational structure provides a framework for establishing and maintaining the long-term support required for successful enterprise GIS operations. This basic organization structure can be useful in managing GIS in small to large organizations, and the same type of organizational structure can be effective in managing community GIS operations.
Identify Key Staff Functions
A broad range of technical skills are required to build and maintain effective enterprise GIS operations. Key operational staffs participate in data management, planning and analysis, field mobility, and operational awareness disciplines. Key supporting positions include skills in application development and enterprise system administration. Figure 11-5 provides an overview of the variety of GIS workflows and key functional responsibilities required to support enterprise GIS operations.
The complexity of these responsibilities will vary with the size and extent of each individual GIS implementation, although every organization will need some level of support and expertise in each of these areas.
Building Qualified Staff
Training is one way to build qualified staff and improve GIS user productivity. Business organizations should develop and maintain a comprehensive training plan to make sure their teams have the resources and skills they need to be effective in their job.
The Esri educational services team provides a Comprehensive Training Plan for ArcGIS for use in developing a training strategy for your organization. The plan includes an overview of the common applications of GIS, the Esri training models, recommended courses for GIS workflows, and a sample GIS staffing plan. This is a great resource for building qualified staff and developing a comprehensive GIS training plan for your organization.
System Architecture Deployment Strategy
Planning is the first step in supporting a successful system deployment. A system design team should review current GIS and hardware system technology, review user requirements, and establish a system architecture design based on user workflow needs. A deployment schedule, as shown in Figure 11-6, should be developed to identify overall implementation objectives.
Phased implementation strategies can significantly reduce implementation risk. Computer technology continues to evolve at a remarkable pace. Integration standards are constantly changing with technology and, at times, may not be ready to support immediate system deployment needs. New ideas are introduced into the market place every day, and a relatively small number of these ideas develop into dependable long-term product solutions. The following best practices are recommended to support a successful enterprise GIS implementation.
- Represent all critical hardware components planned for the final system solution.
- Use proven low-risk technical solutions to support full implementation.
- Include test efforts to reduce uncertainty and implementation risk.
- Qualify hardware solutions for initial production phase.
Initial Production Phase
- Do not begin until final acceptance of pilot phase.
- Deploy initial production environment.
- Use technical solutions qualified during the pilot phase.
- Demonstrate early success and payoff of the GIS solution.
- Validate organizational readiness and support capabilities.
- Validate initial training programs and user operations.
- Qualify advanced solutions for final implementation.
Final Implementation Phase
- Do not begin until final acceptance of initial production phase.
- Plan a phased rollout with reasonable slack for resolving problems.
- Use technical solutions qualified during previous phases.
- Prioritize rollout timelines to support early success.
Implementation strategies are accellerating with faster technology evolution
- Production upgrades are scheduled to enable required technology advancement.
- Product upgrade deployments are integrated into production rollout schedule when ready.
- Enterprise GIS environments are upgraded incrementally to meet operational requirements.
- Functional and performance testing is completed before production rollout.
- Configuration control for each upgrade is critical for implementation success.
Virtual Desktop and Server Technology
Virtual server technology is continuing to mature and reduce the cost of managing a rapidly changing IT environment. Figure 11-8 identifies a deployment strategy taking advantage of virtual server technology. Many GIS operations are being deployed into virtual server environments. Many ESRI development and testing operations are currently supported in virtual desktop or virtual server environments. Vendors are improving management and performance monitoring of virtual server environments, and it is becoming more practical to manage and deploy production environments in virtual server deployments.
Platform virtualization technology provides IT managers with a way to abstract the installed platform software environment from the physical platform hardware. There are two fundamental levels of virtualization, one being virtual desktop environments hosted within a physical platform operating system that interfaces with the physical platform hardware and the other being a virtual server environment hosted on a hyper-visor layer that interfaces with the assigned physical platform hardware. In both cases, the virtual desktop or server contains its own dedicated operating system and software install separate from other virtual systems deployed on the same hardware.
There are many recognized benefits and some potential disadvantages with virtual desktop/server deployments.
- The benefits include faster provisioning times, physical server platform consolidation, fast recovery from system failures, simplified production delivery and recovery, and optimum configuration control. All of these benefits directly contribute to lower overall systems management costs and a more stable operating environment.
- The disadvantages include additional software cost and some performance overhead. There may also be functional limitations (limited access to hardware graphic cards and performance monitoring software) which in many cases can be managed with the proper deployment selections.
The real need for more rapid adaptive deployment schedules (to keep pace with changing technology) which also must be coupled with more stable production deployments (reduced production downtime and more rapid failure recover) drive the need for virtual platform environments - virtualization is one solution that addresses some real IT management needs.
The potential disadvantages can be managed by proper deployment strategies. The performance overhead for virtual desktop environments is much higher than for server environments, and for this reason virtual desktops are normally limited to software development environments where performance and scalability is not a critical factor.
Server consolidation benefits can be leveraged in a Staging environment, were multiple release candidates can undergo test and validation in preparation for production deployment. Several production release candidates can be testing in parallel on the same physical server platform.
Production deployment can benefit from deploying an existing virtual server install (Staging configuration that has completed final test and acceptance) to a higher capacity production physical server by simply moving the Staging server release to the production platform. If there is a production failure identified after deployment; it is a simple process to move the production environment back to the previous release. Deploying virtual server staging environments to a physical server production environment is also a viable option - ensuring optimum performance and scalability for the production environment. Virtual server migration software is available to accomplish these provisioning tasks during live operations with no production downtime.
Virtual server performance impacts will depend on the workflow environment, and can vary between 20 - 50 percent (and sometimes higher) in the most efficient virtual server configurations. Hardware platform performance has improved over 80 percent within the past two years, which more than overcomes the virtual server processing overhead. The new servers also provide higher capacity (discussed in chapter 9), which opens the door wider for server consolidation benefits.
Virtual server deployments appear to be moving to mainstream IT production environment. The big question is no longer whether it makes sense to deploy on virtual servers, but rather when and which software vendor solution will provide the highest return on investment.
Virtualization deployment options
Figure 11-8 provides an overview of the four principle virtualization solutions. This provides a variety of virtualization options available for ArcGIS deployments.
Physical server configuration:
- Desktop applications share a common Server operating system.
- Multiple Desktop applications are virtualized and presented within remote terminal client.
- Windows: AppV (Softricity)
- Citrix: XenApp
- VMware: ThinApp
Esri certifies each ArcGIS for Desktop release with Citrix XenApp server environment.
Best practice: Large number of Esri customers use Citrix XenApp presentation servers for remote user access to centrally managed virtual presentation servers.
Physical server configuration:
- Virtual servers share a common hypervisor for physical server access.
- Virtual server machines have their own installed Server operating system.
- Server applications are installed in each separate virtual server.
- Windows: HyperV
- Citrix: XenServer
- VMWare: VMWare Server
Best practice: Large number of Esri clients deploy ArcGIS for Server in the three available virtual server environments.
Physical server configuration:
- Virtual desktops share a common hypervisor for physical server access.
- Virtual desktop machines have their own installed Desktop operating system.
- Desktop applications are installed in each separate virtual desktop.
- Windows: Virtual PC
- Citrix: XenDesktop
- VMWare: VMWare VDI
Best practice: Virtual desktops are used by Esri for development and hosted ReadyTech training environments.
Virtual desktop environments to not have access to hardware video cards which contributes to some functional performance degradation noticed by users.
Physical desktop configuration:
- Virtual desktops interface through a hypervisor for physical desktop access.
- Virtual desktop machines have their own installed Desktop operating system.
- Desktop applications are installed in each separate virtual desktop.
- Windows: Virtual PC
- Citrix: XenDesktop
- VMWare: VMWare VDI
Best practice: Virtual desktops are used by Esri for development and classroom training environments.
Warning: Some virtual desktop environments to not have access to hardware video cards which contributes to some functional performance degradation noticed by users.
Technology Product Life Cycle
Figure 11-9 provides an overview of the technology product life cycle, from initial introduction of a new idea (product innovation) through end of life. Technology is changing faster every year, and managing technology change within a production environment is a challenge for GIS managers and IT administrators.
We are seeing an increasing number of new ideas introduced into the marketplace, with each idea promising improved user productivity and simplified system administration. These new ideas must integrate with existing systems that are constantly changing, and initial implementation can be painful.
Some of these ideas deliver on their promises, and in time they provide significant productivity advantages and reduce overall cost of administration. Soon a new idea comes along that performs better at reduced cost, and organizations must move on to new frontiers leaving legacy systems behind.
Selecting the right technology at the right time will optimize business performance. Introducing new technology before it is ready for prime time can reduce productivity and increase implementation cost. Delaying too long can result in missed opportunities. Getting the timing right promotes success.
Conducting proper testing at the right time can contribute to implementation success. Functional component and system integration testing should be conducted for new technology during prototype development and before introduction into production. The primary focus during this testing is to make sure everything works. Performance targets established during the initial system design can be evaluated during early testing, paying close attention to map display performance and layer complexity (see Chapter 3 Software Performance). This is an opportunity to evaluate workflow functions and reduce processing overhead.
Enterprise system environments are becoming more complex. Testing should be conducted in a production software environment (same operating system, service pacts, software architecture, etc). Configuration challenges such as firewall access, security, and high availability should be configured and tested before deployment. Development and test environments should be established to represent the complete software configuration for each production release cycle.
Figure 11-10 identifies best practices for planning and conducting functional system testing. Conducting proper testing at the right time can contribute to implementation success. Functional component and system integration testing should be conducted for new technology during prototype development and before introduction into production.
A test plan should be developed with clearly defined test requirements, establish configuration control (software versions, operating system environment), and provide test procedures. Testing should be completed before production deployment. Testing should be conducted using the software versions and operating system that will be deployed in the production environment.
- Complete a risk analysis: Identify new functionality that requires testing.
- Identify test objectives and establish configuration control plan.
- Identify test hardware and software configuration.
- Develop test procedures.
- Identify implementation team and establish implementation schedule.
- Order hardware and software and publish installation plan.
- Conduct test plan and validate functional acceptance.
- Collect test performance parameters (CPU, memory, network traffic, etc.).
Test results and documentation:
- Document the results of the testing.
- Include specific hardware, software, and network components that were tested.
- Include installation and test procedures that were followed, test anomalies, and final resolution.
- Complete test compliance matrix identifying validation of functional requirements.
- Publish the test results and make available for reference during system implementation.
- Complete prototype integration testing before production deployment.
- Test in a production environment (configuration control).
- Document functional requirements and test procedures.
Figure 11-11 identifies some best practices associated with system performance testing.
Performance testing can be expensive and the results misleading. Normally initial system deployments need to be tuned and optimized to achieve final performance goals. Often system performance bottlenecks are identified and resolved during initial deployment. Early application development focuses primarily on functional requirements, and performance tuning is not complete until the final release. Actual user workflow environments are difficult to simulate, and test environments seldom replicate normal enterprise operations.
The scientific method introduced with grade school science fair projects provides some fundamental best practices that directly apply to system performance testing. Performance testing should only be conducted to validate a hypothesis (something you think you know). The primary objective of a performance test is to validate the hypothesis (confirm what you know). The test is a success only if it proves the hypothesis (testing does not teach you what you don't know).
Initial performance testing often fails to support the test hypothesis. With further analysis and investigation, test bottlenecks and/or improper assumptions are identified that change the test results. Performance testing is only successful if it validates the test hypothesis. A scientific approach to testing (Scientific Method) can help develop a true interpretation of the technology. Technology is changing with each service pack release, and this requires an open mind and willingness to continually change what we believe to be true.
System performance testing is best conducted during the initial production deployment. During this phase, real users doing real workflows can generate a real user environment. Critical system components should be monitored during the initial deployment to identify processing bottlenecks and resolve system conflicts. Compliance with design performance targets can be validated by observing system throughput and utilization during initial user operations. Initial deployment acceptance should include validation that user workflow performance goals are met.
Systems Integration Management
Basic project management practices promote implementation success. Project teams should be established, individuals should be assigned specific responsibilities, a task plan should be developed to support implementation planning, a configuration control plan and change control process should be established, and an implementation schedule should be published to support project deployment milestones.
A system architecture design can provide the framework for establishing an implementation plan. The implementation plan should be developed after final selection of the hardware vendor solution. Figure 11-12 provides a typical system deployment schedule. Specific decision milestones should be included in the schedule and each major task effort clearly identified.
An implementation project manager should be assigned to make sure all tasks are well-defined, and every participant has a clear understanding of his/her responsibilities. A clear set of acceptance criteria should be developed for each implementation task and a formal acceptance process followed to ensure integration issues are identified and resolved at the earliest opportunity.
The capacity planning tool can be used by project managers to establish performance milestones and validate performance targets are met throughout deployment. Peak users can be identified for different project milestones, and system throughput and server utilization can be reported at each milestone to demonstrate performance goals are met.
Figure 11-13 shows some of the ArcGIS Server and system tools used for performance monitoring. Additional performance validation tools were discussing in Chapter 3 (ArcGIS Server 9.3.1. map optimization tools and the mxdperfstat performance monitoring tools). The performance terms and relationships discussed in Chapter 5, particularly the relationship between throughput (peak users or peak transaction rates) and utilization (server CPU or network bandwidth utilization), can identify if the deployed solution is performing within the initial project performance milestone.
Monitoring live performance metrics can provide excellent validation that the system environment is designed to support peak throughput loads. The challenge is to collect appropriate throughput and utilization metrics that represent actual business workflow loads (what are the current system loads).
Department (Business) managers can often identify the percentage of peak user loads (throughput) generated on the system during high volume events – these are the same high volume events that drive staffing and business planning needs. System performance monitoring tools can measure and report platform CPU utilization and network traffic during these high volume events. The problem is that business management and IT administration staff seldom share these metrics. These live performance measures are the most accurate resource for understanding whether the existing system environment has the capacity to meet expected peak system loads.
Several third party products are available for system tuning and Web load testing. This presentation by our Enterprise test team shares our system tuning and testing experience and best practices. A variety of typical benchmark test results are published on our Enterprise Resource Center.
Figure 11-14 shows the importance of establishing performance milestones and validating performance during implementation. Measuring display rendering time when authoring a service is the first opportunity to validate performance. Measuring deployed service rendering time is another opportunity to validate performance. The CPT test tab is designed to translate throughput and utilization measurements to workflow service times that can be compared directly to the initial project workflow performance targets. Monitoring progress in meeting performance milestones can reduce deployment risk and ensure project delivery success.
When performance issues are identified early in deployment, proper adjustments can be made before impacting production workflow productivity (simpler map displays - less layers or generalize layers with large number of features, reduce number of current batch jobs during peak system loads, evaluate preprocessing alternatives (map cache, generalized geodatabase layers, etc). Chapter 3 identified a variety of ways to improve GIS display performance. Identifying and resolving performance issues before they become production level performance problems will promote deployment success.
System tuning is a critical part of final system integration and deployment. Initial system deployment is the first opportunity to begin performance tuning. Heavy batch processing efforts should be identified and separated from interactive user workflows and supported through a separate batch process queue. System backups and heavy processing workloads should be scheduled during off-peak workflow periods.
System component performance metrics should be monitored on a periodic basis particularly during peak workflow periods to identify performance bottlenecks and address system deficiencies. Figure 11-15 provides an overview of the components supporting an enterprise GIS environment. Any component has the potential to introduce a weak link in the overall system performance equation.
Managing Technology Change
Enterprise GIS operations require a combination of strategic planning and a continued investment in technology. Technology is changing very rapidly, and organizations that fail to manage this change fall behind in productivity and overall cost management. Managing technology change is a major IT challenge. Figure 11-16 identifies a conceptual system architecture planning and deployment strategy for technology change management.
Enterprise operations should include a periodic cycle of coordinated production system updates.
- Planning, test, and technology validation should occur one periodic cycle ahead of each production deployment.
- Production deployment efforts should be coordinated to support operational technology needs.
Planning, test, and validation:
- Planning activities should be established on a periodic cycle.
- Coordinated to support the organization's operational and budget planning needs.
- Strategic plans should be updated to support a multiyear deployment strategy and published periodically (normally on an annual cycle).
The planning and evaluation process should include:
- A requirements evaluation (strategic plan update).
- Technology refresh (training and research).
- Requirements analysis (process and requirements review).
- Test and evaluation (evaluate new technology alternatives).
- Prototype validation (pilot test programs).
Best practice: Efforts should be scheduled to support the annual system deployment upgrade cycle.
- Operational system upgrades should be planned on a periodic cycle.
- Scheduled to implement validated operational enhancements from the planning and evaluation program.
- System deployment phases should include initial implementation (implementing changes in an operational test environment) to support deployment authorization.
- The program should also include planned schedules for new technology procurement and deployment on a periodic schedule (in some cases deployment upgrades can be implemented on a monthly or quarterly basis).
- All production system upgrades should be planned and scheduled with full support for ongoing operations.
Successful implementation depends on a good solid design, appropriate hardware and software product selection, successful systems integration, and careful incremental evaluation during installation. A phased approach to implementation reduces project risk and promotes success, providing the opportunity for early success and flexibility to incorporate new technology at low risk prior to final system delivery.
Guidelines are available to support a successful system design, even for large complex systems. Final purchase decisions are influenced by both operational requirements and budget limitations, introducing unique challenges for system design. Good leadership, qualified staff, and proven standard practices support successful deployments.