that can easily be calculated and related in hard-dollar amounts. Taking a server down on an Internet site is like closing the physical doors to a store, which would send your customers to the competition. Cyberstores cannot afford that.
Every server administrator thus needs to zero in on this subject with the determination to maintain high availability of servers and services at all times. In fact, all services and components in Windows 2003 should be listed on an availability chart or a risk-assessment chart. The following list enumerates several areas that have been built by Microsoft with high availability criteria on the agenda:
♦ Bouncing server syndrome
♦ Clustering and server redundancy
♦ Storage redundancy
♦ Disaster recovery
Bouncing server syndrome
We're not sure who first used the word bounce to refer to the act of rebooting a server, but the term has caused many good laughs. In mid-1999, we were joined by a good-natured but overly serious Syrian-born VMS administrator who took over the management of a nationwide network of DEC VMS machines. One day, he came over to network administration and told us that he had just received the strangest call from one of the remote centers. "They say that they need me to bounce the Coral Gables VMS, but this is the first time I have heard such a term." We replied, "Yes, you need to pick it up and drop it on the floor . . . do you need help?"
From that day on, our VMX administrator would often tease us about the number of times that you must "bounce" NT (a lot more than a VMX machine). Another term that is often used in IT circles is ¡PL, which stands for Initial Program Load (an operating system restart) and which is rooted in legacy host systems and midrange talk. All systems require you to reboot, bounce, or IPL. Availability is rated according to how often you must reboot.
Windows NT has a horrible availability rating. Just about any configuration change that you make demands a reboot. If you have administered NT for any number of years, you know that you just need to open the network configurations utilities and look at the settings and be told that you must reboot. Often, we would just ignore the warning and hit Cancel. Far too many changes, however, require you to reboot an NT server. At times, we wondered whether just breathing on the monitor would require a bounce.
Microsoft has improved the reboot record in the Windows 2003 kernel tremendously, both for new services and situations where applications and services crash. Improvement is especially noticeable in areas where you typically make a lot of changes, such as the network configuration and so on. Static IP address changes are, for example, immediate, as are reconfiguring network interface cards (NICs) and the like. A number of areas still need to be improved, such as installing software (such as service packs). Although a service pack reboot may be forgiven, reboots while installing new user applications on the server devoted to terminal users is not. A reboot after promoting a domain controller, however, is understandable. Still, it is hoped that later versions of Windows Server 2003 require even fewer reboots.
Windows 2003 Enterprise Server now has clustering services built in, which is a big improvement over the Cluster Server product that was shipped as an add-on to Windows NT and Windows 2000. Clustering is a form of fault tolerance that enables users connected to one server to automatically connect to another server if the former server fails. This situation is technically known as a failover in clustering terms. (We do not deal with clustering in this book because our scope of coverage is limited to Windows Server 2003.)
Clustering, however, is not only applicable to redundancy, but also to load balancing, and particularly network load balancing, which clusters network resources. With technologies such as IntelliMirror and Group Policy, your users need never know which server in a farm of 50 machines is currently servicing their needs. The distributed files system, folder redirection, offline files and folders, and more all play their part in clustering and availability.
Storage services in Windows 2003 play a critical part in availability. Windows 2003 supports all available levels of disk arrays. The distributed file systems and NTFS 5.0 have several key features that support the high-availability initiative.
R2 provides Storage Manager for SANs, which is a software feature that administrators can use to create and manage the logical unit numbers (LUNs) that are used to allocate space on storage arrays in both Fibre Channel and iSCSl environments. Administered through a conventional snap-in, Storage Manager for SANs can be used on storage area network (SAN)-based storage arrays that support Virtual Disk Server (VDS) using a hardware VDS provider. Because of hardware, protocol, transport layer, and security differences, configuration and LUN management differ for the two types of supported environments.
Disaster recovery is managed by using Windows 2003 remote and removable storage services to maintain reliable backup sets. The System Recovery Console enables you to boot to an NTFS-supported command line that enables you to access NTFS volumes. ln addition, Windows 2003 also boots to a menu of "safe mode" options in the event of serious system instability, and gives you the opportunity to provide a description of why the system went down for logging purposes.
Security services are critical to Windows 2003. In our opinion, you cannot have enough tools to protect the network. We discuss this topic further in the following section, in depth in Chapter 3, and in places in this book where security configuration is required. You would do well to take stock of the thousands of hacker sites on the Internet and join as many security groups as possible. The age of e-terrorism is upon us, and Windows Server 2003 systems are among the main targets.
Was this article helpful?