F E A T U R E

Endurance 4000 Goes the Distance

November 29, 1999
By Stacy Morang

In the worlds of finance and manufacturing-process control, every minute a server isn't working costs thousands of dollars. Even mundane applications such as e-mail are critical. It's highly unlikely the designers of the IBM PC ever dreamed that their design would be used for high availability, and yet now it is. RAID and redundant power supplies have been added to plug some of the bigger holes, but hefty cracks remain.

Several software-based assured-availability systems offer a solution; these usually require specially written applications and some time to start up a new instance of the failed application. Stratus Computer, Sun Microsystems and Tandem (a division of Compaq Computer Corp.) provide hardware-based solutions, but to date none have a Microsoft Windows NT version, though Stratus has plans for a Merced-based system.

Now, along comes Marathon Technologies' Endurance 4000 to address Windows NT high availability. We had an opportunity to put the Endurance system through its paces and found it to be solid. We liked its simplicity, ease of management and the simple fact that it works. Unfortunately, if you need multiple processors or load-sharing you'll have to look elsewhere (or wait for the symmetric multiprocessing version to arrive). Still, the Endurance 4000 is an excellent solution: Its single processors are fast enough for many jobs and the worry-free system will ease the burden on your IT staff.

Hardware-Software Combo
Endurance is hardware and software designed to combine four personal computers into a single Windows NT server with no single point of hardware failure. And, almost more important to us, the software applications need not be aware that they are running on a high-availability server. In fact, if you're willing to work with an 800 x 600 screen in 16 colors, you could have a high-availability workstation--regular applications work just as they always have.

There are several advantages to fault tolerance in hardware over software-based systems. Programs don't need modification to be fault-tolerant, which is a substantial advantage for shops with custom applications. The failover time is less than it is with software-based failover, and recovery is handled automatically once the failing parts are repaired.


In the Endurance 4000's setup, the computers in the system are divided into pairs, or a tuple in Marathon's lingo. These tuples can be separated by up to 1.5 kilometers for power and space redundancy, which is a great feature for those of us working in basement computer rooms near fragile old plumbing. This setup lends itself to a range of vital 24 x 7 functions, from process control to the Web server that is your primary communication link with clients or staff (see "Logical Hardware Interconnection for the Endurance 4000," above).

If you don't buy the turnkey version of the Endurance system, you must supply the PCs that house the Marathon interfaces (see www.networkcomputing.com/1024/1024f4side2.html for more about how the Endurance 4000 works). As you might expect, some hardware is incompatible with the Endurance architecture, so you must verify your choices with the company. Marathon supplies all the hardware, cabling and software needed to blend the four computers into one. With the software comes an extensive online manual that makes fault resolution possible even for someone with only a general understanding of the system.

Managing Your Endurance
Management is achieved through a basic but fully functional real-time display GUI or through a set of scriptable commands for those of you who harbor a certain fondness for the text terminals of yore. Simple but effective, the colors of the icons change in real time to reflect the state of the system, and double-clicking lets you select from all the commands supported by that device. To round things out, several menu entries are provided that offer shortcuts to often-used NT programs, such as the Event Log Viewer.

The GUI program can be run on any Windows NT system connected to the same network (just in case you don't want to live in the machine room). The command-line interface commands have all the functionality of the GUI and a bit more. They're handy for forcing the system to do something automatically at a specific time, in concert with the NT Scheduler Service, for example.

We liked Marathon's open attitude to systems management very much--the vendor is willing to admit it does not know the "one true way" to manage a system, a refreshing change from those vendors who seem so narrowly focused on their own particular view of the world.

The Endurance 4000 is around the top of the price chart for a high-availability NT system, so you will need to calculate just how much downtime will cost you. The alternative is software-based clustering, such as that offered by Legato Systems' Octopus HA+, Microsoft's Cluster Server, NCR Corp.'s LifeKeeper and Vinca Systems' Co-Standby Server. These vary considerably in price. While they generally require more expensive hardware because RAID arrays and other redundancies are more important, in the end, software clustering should be somewhat less costly initially. But while you save money, you'll lose the virtually instant failover and the ability to use any software, not just cluster-aware applications.

If your office is at all like ours, e-mail failure alone is often a sufficient excuse for extensive staff testing of the Windows Solitaire program--and it isn't cheap for the IT staff to solve the problem.

Also, consider your budget for those things your computer room absolutely must have. Don't even bother with an Endurance system if you're unwilling to buy or can't afford to budget for decent hardware--along with reliable server boxes, a UPS is required (and in fact, having two is preferred), and you'll still need good backup equipment and appropriate software because all the hardware in the world can't protect you from an application bug or user error.



PAGE: 1 I 2 I 3 I 4 I 5 I NEXT PAGE
 

Research and Reports

Storage Virtualization Guide
May 2012

Network Computing: May 2012

TechWeb Careers