Top ten server trends take shape

The hardware market is changing fast. With the advent of the Linux and Windows XP operating systems, and continued falling prices for powerful and efficient chips, users are demanding more of their hardware than ever before.

For design engineers, this could mean the freedom to collaborate remotely with colleagues and customers, to share large graphic CAD files over secure lines, or to run more iterations of number-intensive simulations like CFD.

But first they need the hardware to make it happen. According to engineers at Sun Microsystems, this new tide of computing power will develop in ten main server trends over the rest of 2002.

The top three trends may be emerging the fastest, says Chris Kruell, group marketing manager of Sun's computer systems group. As companies look for ways to reduce total cost of ownership (trend #3) of their computing resources, they try to eliminate redundancies.

In their rush to grow, many organizations now suffer from "server sprawl," so they have stacks of servers sitting in closets somewhere, underutilized and expensive. But Sun's Starfire software uses partitioning to share these scarce resources, enabling IT departments to consolidate (trend #1), using fewer servers for more applications, he says.

This also allows companies to migrate their workloads off expensive mega-mainframes onto several smaller servers (trend #2).

Yet even when they're streamlined, these server farms and mainframes aren't portable. In this age of round-the-clock manufacturing, someone in the company is always working; frustrated they can't tap into unused computers. Sun's solution is Sun Grid Engine (trend #8)-a type of distributed resource management (DRM) software, available free on the company's website.

Microsoft also makes a computational cluster, working with its business partner Intel to provide shared computing power for CAD, PDM (product data management), desktop analysis, and PLM (product lifecycle management), says Chris Ray, Microsoft's global industry director of engineering.

Sun's software allows users in various locations to share power on a central CPU(s). The first step is the cluster grid, used within a department. The next step is a campus grid, so various departments or project teams can share their workstations, servers, data storage, and networking. And finally, a global grid could exist between separate organizations or companies.

In its first 12 months (ending Sept. 2001), new users applied Sun Grid Engine to 118,000 CPUs, nearly all in cluster grids, says Peter Jeffcock, group marketing manager in Sun's volume systems products group.

But industry still has security concerns, he admits. "Even if you use authentication, and the data is secure at both ends, it's still a public network. You also need better bandwidth. We'll be there in 3-5 years, but now most of the activity is at the low end. Sun will be ready when its customers are."

So engineers must squeeze all the power they can from their current hardware. Thus the computing world is creeping toward 64-bit computing (trend #4), slowed by the extra work to compile software and operating systems for the new platform.

"It's crucial when customers have large datasets," says George Iwaki, senior product manager, Sun's technical markets products group. "You get 4 GB addressable memory on a 32-bit chip (2 to the 32^nd), but 18 hexabytes on a 64-bit chip (2 to the 64^th)."

Indeed, Microsoft's Windows XP released first in 32-bit format, with 64-bit to follow. And PTC recently released a 64-bit version of Pro/ENGINEER.