Other Considerations in Data Center Design

There are a number of issues that may apply to a company when considering the centralization of its MetaFrame servers. It is not possible to anticipate every conceivable issue of designing a data center, but the following topics cover some issues we have run into in the past that may help in planning.

Legacy Hosting

Determine if applications will run on MetaFrame servers that need to access data or programs on legacy systems (enterprise resource planning (ERP), database query and reporting tools, and terminal emulation are all examples of such applications). If this is the case, the legacy systems and MetaFrame servers should be co-located to optimize the network bandwidth required between these systems, as shown in Figure 5-1.

Figure 5-1: Legacy systems located near MetaFrame servers

Offsite Data Storage

Even with a secure and reliable data center, data backups should be taken off-site to a hardened location or copied to an off-site location daily. A hardened location is one in which proper fire and moisture protection has been ensured, as well as physical security for data storage media. During a production failure or disaster involving a loss of site, such backups can mean the difference between a quick recovery and no recovery at all. Many national and regional firms specialize in data storage. Other firms will use a frame relay connection or the Internet to back up data to a secure offsite location. If a company is outsourcing its data center, they must make sure they have tape exchange or electronic vaulting as part of the service level agreement (SLA) with their vendor. Otherwise, one of their own people will have to travel to the data center daily to change tapes.

Unusual Connectivity

An organization needs to consider if they have systems with network topologies or connection requirements different from those of their MetaFrame servers. Any different topologies, such as SNA, token ring, serial lines, and telephony, will need to be incorporated into the network and facilities design. Bridging or conversion technologies may be required, and they may not work the same way in a MetaFrame and Terminal Services environment as they do in a standard Windows environment. For example, if there is a Computer Telephony Integration (CTI) application that allows users to retrieve their voice mail through a PC Windows interface utilizing individual IP addresses, it will probably not work on a MetaFrame XP server without modification since a MetaFrame XP server will use the same IP address for all users.

Nonstandard Systems

Card readers, document scanners, time-card systems, or other automated systems with dependent applications must be taken into account when building a MetaFrame server farm. Depending on the exact nature of these automated systems, one may not be able to incorporate them as part of the server-based computing architecture. Here are some guidelines for making sure these systems will work in a new environment:

• If the system has associated software that runs in DOS, see whether there is a Windows 32-bit version. Even better, look for a version that has been tested and certified with Terminal Services. Windows 2000 and Windows 2003 no longer support DOS applications running on Terminal Services (although some DOS applications still work).

• If the system has code that already runs on a server (such as NetWare or Windows 2000), see whether you can keep the server in place and run the client software on the MetaFrame server.

• If the system runs at the user's desktop, make sure any services it needs, such as printing or use of serial ports, will work with Windows Server 2003 or MetaFrame XP's port redirection capabilities.

• Test these systems sooner rather than later in the deployment cycle so that there is time to respond if an upgrade or complete revision of the design is needed to find a new solution.

Rogue Servers and Applications

An organization should ask itself if its group or project team is in control of all the servers in the enterprise that may be affected by a project. Especially in a large enterprise, it is likely that some servers and applications have been set up regionally without their knowledge. Unless it actively investigates beforehand, the first time a company hears of such systems may be when they disable a network circuit or otherwise cut off the regional users from the rest of the network. It is wise to develop a plan to have a sunset period in which these locations are given a certain amount of time to phase out these systems and begin to access their applications from the new data center.