The Project Design Plan

The project plan incorporates all aspects of the SBC design. This plan includes both the project definition document and results of the infrastructure assessment. The financial analysis performed by the feasibility committee should be fine-tuned throughout the planning process until the final planning document includes a solid estimate for project costs.

The planning document should clearly convey the organization's server-based computing migration strategy and be suitable for presentation to both executives and auditors. It discusses the various options that the planning team considered for each major component of the project, along with the rationale behind the team's ultimate decision.

Incorporated as part of the plan should be the considerations given to capacity planning, redundancy, and security. The tremendous advantages accruing from the centralization that SBC enables can become huge liabilities if scalability is not easily incorporated, a crucial data center component without a redundant part should fail, or the system's security is compromised.

The project plan also serves as a roadmap for the project managers and implementation team as they work to institute an enterprise SBC environment. Detailed explanations of the design plan are discussed in the remaining chapters in this part of the book. An overview of the design plan follows.

Terminal Services Design

Designing the Terminal Services environment will be difficult for organizations unfamiliar with the basics of server-based computing. We recommend that the appropriate IT personnel take courses in Terminal Services, MetaFrame XP Presentation Server, and Advanced MetaFrame XP Presentation Server before beginning the design process. Since SBC consolidates processing into a central data center(s), it is important to plan for redundancy of all key components. The MetaFrame XP Presentation Server server farm with load-balancing has built-in redundancy, but special awareness should be given to implementing it for other servers such as Windows Server 2003, SQL, e-mail, and web services. Redundancy should also be included for other critical components such as network switches, load-balancers, routers, storage units, and bandwidth management devices.

Application Architecture Design

Define the strategy both for deploying the on-demand enterprise environment and for handling legacy applications that will not be supported. Users may be allowed, for instance, to run legacy applications locally as long as they want. Alternatively, they may be given a deadline for transitioning to corporate-approved and -supported applications. (Application strategies are covered in more detail in Chapter 13.)

Data Center Architecture Design

The planning team needs to determine the number of data centers, based upon demographic, geographic, disaster recovery, and business requirements. They must evaluate site considerations, including power, cooling, fire suppression, and physical security. They also must evaluate options for either hosting the centers internally or using co-location centers, such as AT&T or Verio. (Data center architecture is discussed more thoroughly in Chapter 5.)

Disaster Recovery/Business Continuity Design

Since users are completely dependent upon SBC for all of their hosted application processing, disaster recovery is an extremely important component. Fortunately, server-based computing makes a real-time disaster recovery solution far more practical and affordable than in a PC-based computing environment. (A variety of disaster recovery options are discussed in Chapter 19.)

Network Backbone Design

Each data center requires a high-speed backbone connecting the MetaFrame server farm with other servers in the data center. Small organizations may be able to get by with 100-Mbps switched Ethernet. Large firms will likely require ATM or switched gigabit. Redundant network interface cards (NICs) and switches should be incorporated as part of the design. (The network backbone is discussed more thoroughly in Chapter 6.)

Server Farm Architecture Design

The findings from the proof-of-concept pilot program will enable the planning committee to select server quantity, type, sizing, and configuration for the Terminal Services implementation. The Citrix resource manager component and Windows Performance Monitor can help determine server scalability. Other tools can simulate server loads. If DOS or 16-bit applications will be run, extra servers may be required for a multitiered server farm. Likewise, support for multiple languages will necessitate additional servers for each language. (Server farm architecture is discussed more thoroughly in Chapter 12.)

File Services Design

When users store all of their data at corporate data centers, unique problems arise in handling file services efficiently. The project team should evaluate the different options, including server clustering of general-purpose file servers, storage area networks (SANs), and network attached storage (NAS). Archive systems and backup software and services must also be selected. (File services are covered in detail in Chapter 6 and in Appendix A.)

Print Server Architecture Design

Printing tends to be one of the most difficult and time-consuming parts of an enterprise server-based computing implementation. Decisions must be made about the configuration of one or more central print servers at each data center as well as the type and quantity of print servers in remote offices. Just a few of the other printer-related decisions the team will have to make include integration of host system printing, local PC printing, printer auto-creation to create temporary printer assignments for mobile users, trusted print sources, lockdown of registries, and control over printer access. (Printing is discussed in detail in Chapter 18.)

User Profiles Design

Most Terminal Services implementations utilize either mandatory or roaming profiles, but we recommend that organizations use scripting to enable desktop lockdown while allowing users the flexibility to select default drives and printers. (We present our scripting techniques in Chapter 15.)

Login Script Design

To minimize administration, there should be one script that works for both fat and thin clients. Additionally, login scripts should be designed to run very quickly and efficiently.

Automation Design

You will want to automate application installation and updates, server imaging processes, and client installations using products such as Citrix Installation Manager (IM) and Norton Ghost. (Automation design is covered exhaustively in Part III of this book.)

Windows Server Design

Designing a Windows server infrastructure to support Terminal Services is a key part of the planning process. The following components are included.

Domain Model Design

When designing large enterprise infrastructures, it is easy to design a domain that entails frequent and inefficient replication. For a server-based computing environment, we generally recommend a single master domain model to separate authentication of users and groups from authentication of resources. (Domains are explained in more detail in Chapter 8.)

Active Directory Design

Active Directory planning and configuration tends to be much simpler in an enterprise ODE environment. This is because there is generally no need to worry about intersite replication since all domain controllers are in the data center. Another concern that is eliminated is the need to accommodate authentication over the WAN by users in small offices without local domain controllers. (Active Directory Design is covered in Chapter 6.)

WINS Architecture Design

Under server-based computing, WINS is less likely to be required. (In Chapter 6, we discuss when and how to use WINS.)

Backup Architecture Design

If the infrastructure assessment reveals inadequate backup systems to handle the demands of centralized data storage, the archive systems and backup software and services require selection. (Backup systems are covered in Appendix A.)

Back-End Database Design

The size and configuration of a back-end database in a server-based computing environment, where all users will be hitting one database at one time, will often be different from a distributed database model, where several database servers are located across the enterprise. The distributed servers would handle a relatively small number of users and have replicated data backed up and stored at a central point. The server-based computing model might require far more powerful database server(s) or clustered servers at the data center, depending on usage, as well as middleware application changes.

Network Design

A sound network infrastructure is vital to supporting an ODE environment. In addition to remedying any shortfalls discovered during the infrastructure assessment, the following issues should be addressed.

Topologies

Because the MetaFrame ICA Client uses such little bandwidth, the composition of the topology to the desktop is generally not of great importance as long as it is reliable. In most cases, 4-Mbps Token Ring will deliver the same performance as switched 100 Mbps.

WAN Architecture

An ODE environment requires a robust, scalable, and highly reliable WAN design because remote office users are completely dependent upon the MetaFrame servers at the corporate data centers. The planning team must evaluate the different connectivity options, including the Internet and redundancy options. During the transition from PC-based to server-based computing, residual traffic will chew up an inordinate amount of bandwidth. The project plan must allow for this temporary increased bandwidth requirement during the migration process. (Bandwidth management, including packet prioritization, is often essential in order to ensure adequate performance in a Terminal Services wide area network.)

Alternative bandwidth capabilities should also be designed into the system. For instance, if the primary connectivity to remote offices is frame relay, alternative DSL connections to the Internet should be available for backup. Even a dial-up line can keep people in business in the event of a major catastrophe. The ultimate redundancy is to utilize the unique capabilities of the ODE to build a disaster recovery solution with multiple fail-over data centers. (Wide area connectivity is discussed more thoroughly in Chapter 5.)

Remote Access Design

The project planning team needs to choose the appropriate remote access strategy, whether using leased lines, frame relay, dial-up lines, or the Internet. (Remote access design considerations, including connection and redundancy strategies, profiles, and gateway routing, are detailed in Chapter 6.)

The Systems Management Environment

If the infrastructure assessment indicates that a network management package is already utilized as part of the existing PC-based computing environment, the planning team should extend it to encompass the SBC architecture. The team should also decide on how the existing network management package, or a new one, can best be configured to work with server-based computing packages such as Citrix resource manager. (Network management environments are covered in detail in Chapter 9.)

Metrics Design As part of the systems management environment, the planning team should determine which metrics are to be collected and analyzed in order to develop strategies for expansion and for limiting bottlenecks. Citrix resource manager is a good tool to use in this capacity, though it may overlap with existing utilities such as HP OpenView.

Policies and Procedures Design

As is the case with the mainframe model of computing, clearly defined policies and procedures are essential for enterprise SBC success. Adding an application or making a small change to a central router can have severe consequences for hundreds or thousands of SBC users. Although we continue to emphasize the numerous advantages of an enterprise server-based computing environment, it does require that the days of the network cowboy come to an end.

Having been raised, from an MIS perspective, in the midrange and micro eras of computing, my staff had a hard time rethinking the way they do things. For example, while "maintenance windows" were commonplace in the days of the mainframe, they've seemingly disappeared in the PC era. My network technicians were used to shutting a system down minutes after announcing it. We all had to relearn what the MIS personnel we replaced 10 to 15 years ago knew as second nature.

—Anthony Lackey, Vice President of MIS, Chief Technology Officer, ABM Industries

Data Center Policies and Procedures Design

The planning document should include the organization's strategy for managing environmental changes.

Tip

Depending upon the current policies and procedures as revealed in the infrastructure assessment, new requirements may be necessary. For example, a workflow-enabled database should track all changes by administrators and implementers to the Terminal Services infrastructure.

User Policies and Procedures Design

Decisions must be made about data access, device access, and adding new devices. For example, will users be allowed to access local devices from a Citrix ICA session? If so, this policy can have unanticipated ramifications, such as security concerns. (Policies and procedures are discussed more thoroughly in Chapter 15.)

Client Design

The planning committee should identify the different client categories and the levels to which they are expected to utilize SBC services. They must further decide how to specifically set up the clients, and how to configure user desktops. Choices must be made regarding policies for local browsing, emulation, drive mappings, PC local operating systems, local hardware peripherals, and integration with handheld devices such as Palm Pilots. If Windows terminals will be used, the planning team must evaluate the different options and choose the brand and models most appropriate for their organization. (Client implementation is discussed more thoroughly in Chapter 7.)

Client Operating Systems

A primary benefit of centralized computing is the standardization of applications. While standard client equipment and operating systems make administration easier, one of the most compelling strengths of server-based computing is its ability to effectively manage a heterogeneous environment. Still, different operating systems do have different ramifications for functionality under Terminal Services.

User Interface Design

Users can launch entire MetaFrame XP Presentation Server desktops, or simply click icons generated through Citrix Program Neighborhood. The Citrix web interface component enables application publishing to a browser. Citrix MetaFrame Secure Access Manager allows access to all features of the web interface component of MetaFrame Presentation Server as well as the ability to aggregate information from across the enterprise, the Internet, and other data sources and to present it to users in a secure, personalized manner.

Integration with Local Devices

Design strategies must be included for client integration with local printers, handheld units, scanners, bar code readers, and cash drawers.

Non-Windows Client Design

MetaFrame XP Presentation Server enables UNIX workstations, as well as Linux and Macintosh users, to run Windows applications without requiring a separate PC. MetaFrame XP Presentation Server for UNIX (UX/AIX/Solaris) adds the functionality of the X-Window protocol. Users can subsequently launch either Windows or UNIX applications from the same screen.

Data Organization Design

When users migrate to Terminal Services, policies will need to be set about where their data will be stored for different applications (central server storage versus local storage). Creating broad policies that extend across all access infrastructure users will greatly facilitate the ability of help desk personnel to provide prompt support.

Client Application Design

Different application strategies may be appropriate for different categories of users. For instance, mobile users will likely have some local applications, while office users may have none.

Other Client Design Considerations

Groups, drive mappings, and login script strategies must be designed for the different categories of users.

Security Design

Although security should permeate all aspects of the project design plan, a specific security strategy should be identified. Firewall integration, account management, auditing, and the Terminal Services registry should all be included. (Security is discussed more thoroughly in Chapter 9.)

General Implementation Design

The implementation plan should cover training, user communications, data migration, project management, change management, and customer care.

Training Plan

A training plan needs to be designed for support personnel, system administrators, and end users.

Tip

Once end users are set up to access their desktop through Citrix, you can coordinate a more formal introductory training class by using the MetaFrame XP Presentation Server shadowing capabilities. The trainer can have several users simultaneously shadow her PC. Setting up a concurrent conference call provides the audio to describe the visual orientation.

Support Personnel The low administrative requirements of server-based computing combined with features such as shadowing will enable help desk personnel to support many more users once the on-demand enterprise migration is complete. During the transition, however, increased staff and training will likely be necessary to handle the demands of the new architecture while supporting users on the old PC-based computing platform.

End Users Distribution of rainbow packets for general information and at-a-glance documents for frequently asked questions are an expedient way to provide quick user orientation to server-based computing. A rainbow document is modeled after the colorful organizational wall charts found in many hospitals for quick reference to services and locations. The rainbow document literally contains a rainbow of colored sheets, each a bit narrower than the other, providing easy reference to the topics on the exposed edge. Some relevant topics might be "Getting Help," "Finding Your Files," "Glossary of Terms," and "Your Thin-Client Desktop."

Project Management

The planning team should incorporate the essentials of project management as part of the plan. Implementation teams must have well-defined tasks, and required resources must be identified. An estimated timeline for the project beta testing and rollout should be included as part of the planning document.

An enterprise server-based computing migration requires project manager authority, stakeholder buy-in, project reporting and tracking, task assignment, project change control, scope creep control, organizational change management, and timeline management. (Project management is discussed in detail in Chapter 10.)

Change Management

The planning document should include the organization's strategy for managing environmental changes in order to enhance management and end-user benefits. Administrator and end-user training, user reference guides, asset tracking, and a frequently asked questions (FAQs) database should all be incorporated as part of the project. The planning team should include survey forms for gathering information prior to implementation and for measuring user satisfaction as the rollout takes place.

A method for communicating migration plans to users is a very important component of change management. While an on-demand enterprise will provide users with enhanced capabilities and support, it still involves change. Advanced orientation and education will make the process go much more smoothly. (Strategies for internal marketing are discussed in Chapter 10.)

Customer Care

The help desk department will be able to handle many more users once the migration to server-based computing is complete. During the transition, however, increased staff may be necessary to handle the glitches of the new architecture while supporting users on the old PC-based computing platform. (Customer care is given further consideration in Chapter 10.)

Migrating to Server-Based Computing

The planning document should include a roadmap for migrating from fat client to thin client. Also clearly documented should be strategies for consolidating data from both PCs and remote office servers, thus minimizing downtime, and creating a "virtual call center" based upon skill sets.

Expanding the Pilot Test to a Beta

The planning team must decide at what point the proof-of-concept pilot test will be expanded to a beta implementation, and they must decide the parameters of the beta. Objectives should be defined and results measured in order to allow adjustments to the team's migration strategy if required. A scope variance process needs to define who has authority to sign off on out-of-scope items, for example, including a new application as part of the beta. (The beta implementation is discussed in more detail in Chapter 10.)