WorldMark 4300 Terminal Server

NCR's WorldMark 4300 Terminal Server lets IS departments provide continuous computing through thin-client terminals that access a fault-tolerant central storage system. The combination of the WorldMark 4300's hardware; Windows NT Server 4.0, Terminal Server Edition; and NCR's LifeKeeper 2.0 clustering component brings high availability and thin-client computing to the enterprise forefront. I tested a WorldMark 4300 in the Windows NT Magazine Lab to see if NCR's solution successfully implements thin-client technology.

And in This Cabinet
The WorldMark 4300's 77" cabinet houses two 4-way servers and an NCR 6257 disk array subsystem. The two servers are identical in both size and shape. Each server contains four 200MHz Pentium Pro processors, 1GB of memory, a 4.3GB hard disk, and one SMC 10/100 Ethernet adapter. Each server has SCSI connections to the disk array made up of ten 4GB hard disks. Terminal Server and LifeKeeper are installed on both systems, but you must configure them.

The two servers connect to an Apex keyboard/video/mouse switch (KVM switch) that is rack-mounted inside the WorldMark 4300's cabinet. You connect a monitor, keyboard, and mouse to the KVM switch. This configuration lets you toggle between each server's display using the Print Scrn key and lets you use the arrow keys to select each server from a pop-up menu.

New Life with LifeKeeper 2.0
One benefit of the NCR Terminal Server solution is that it brings LifeKeeper to the Terminal Server community. You can use LifeKeeper to achieve the best application availability possible. LifeKeeper lets you depend on maximum application availability during scheduled or unscheduled downtime.

An enhancement to LifeKeeper, known as Cascading Recovery, lets an application endure against multiple failures within a cluster. Thus, an application simply cascades over to another system if one system fails. For a review of LifeKeeper, see "Clustering Software for Your Network," July 1998. For more information about clustering solutions and LifeKeeper, see Mark Smith, "NT Clustering Solutions Are Here," June 1998.

LifeKeeper provides failover capabilities for as many as 16 servers per cluster. This functionality results in higher availability than most NT solutions. LifeKeeper's robust fault-tolerance features provide failover support for the entire server rather than support for only front-end or back-end applications.

LifeKeeper lets users define three components: protected alias host names via its LAN Manager recovery kit, protected IP addresses for continuous access via NCR's TCP/IP recovery kit, and continuous access to volumes for fault-tolerant storage. You can configure these components into a protected resource hierarchy for each logical user group and achieve logical load balancing, to match users to specific Terminal Server hosts. Logical load balancing is different from capacity load balancing, in which users log on and load balancing software provides access via the least busy server.

When a server configured with LifeKeeper goes down, resources move to a different server in the cluster. This migration lets each Terminal Server system provide primary access and fault tolerance in the event of a malfunction. For example, you might have two Terminal Server systems providing thin-client access to two departments. The first department has light users running smaller, code-efficient programs and applications that don't require much memory. The second department has power users running CPU-intensive applications that require more memory and have several applications running at the same time. Users from each department are on separate servers. LifeKeeper exists on both Terminal Server systems, and each user group has a tailored resource hierarchy. If one of the Terminal Server systems encounters a failure, LifeKeeper detects the failure and transfers the hierarchy from the failed server to the running server. When users of the failed server reconnect via normal procedure, they access the failed server's resources; however, the running server automatically services its original users in addition to the failed server's users.