News

Stratus Makes 100% Windows Uptime Guarantee

• By Scott Bekker
• 10/22/2002

Stratus began offering fault-tolerant servers running Windows 2000 Advanced Server last year. The company brought fault-tolerant technology to the Windows and Intel space after about 20 years selling fault-tolerant systems in higher-end areas of the market.

The new program is called the Perfect Performance program. It involves the Stratus ftServer 6500, a four-way server that the company recently introduced to the market. If a customer experiences any hardware or operating system failure in the first year of the contract, Stratus vows to pay $100,000 in cash or product credit.

That payout is what makes the Stratus program ground-breaking, according to Gartner analyst Eric Rocco.

"It's the first announcement we've seen that really offers anything in real dollars for downtime," Rocco says. "Things like this tend to condition the marketplace. It will force customers at the very least to think about asking those kind of questions of the other vendors."

Stratus notes that the system costs one-third less than a typical Unix cluster and one-fifth the cost of an HP (Himalaya) NonStop server. Even so, the guarantee does come at a substantial cost from the perspective of typical customers of Intel-based servers running Windows.

The four-processor ftServer 6500 with an option called Triple Modular Redundancy (TMR), which is required for the guarantee, starts at about $150,000. Customers also must buy a $25,000 to $30,000 service contract on top of the system cost for the guarantee.

Stratus ftServers operate using a concept of modules to provide fault-tolerance. Modules containing a processor and memory simultaneously process transactions in a system. A Dual Modular Redundancy (DMR) system from Stratus has two physical processors and sets of memory cards for every logical processor and set of memory cards in a system. For example, a four-way ftServer 6500 in DMR mode has eight physical processors; in TMR mode it would have 12 physical processors -- with two redundant modules ready to take over transactions if the first module failed.