Designs come full circle

By Graham Prophet, Editor -- EDN, 4/13/2000

For people who believe that everything is cyclical if you wait long enough, the past few weeks have been interesting. At the March Design Automation and Test Europe conference and exhibition in Paris—an event that appears to have come of age and now attracts a

cross section of Europe's EDA-user community and a good representation of EDA vendors—"server farms" were a notable trend in the exhibition hall. Server farms, which essentially are clusters of closely networked workstation processor boards, have been around for a while. Often, they function as a computing resource for jobs typified by long simulation runs that are too large for desktop machines or would take too long to run on a single desktop.

Sun Microsystems' demonstration exemplified this new trend. The company's exhibit included a single-rack cabinet that housed a cluster of processor units. The engineers' desks (or, in this case, the tabletops that stood in for them) contained just a monitor, a keyboard and a mouse, and a network-connection unit. The company calls this setup its "Sun Ray" configuration and touts it as an efficient alternative to having one workstation per desk: When you have a job with a heavy computing load, you can get the multiprocessor resource you need; all of the office applications can run on one of the rack's processors; you get more free desk space because the processor box is gone; and you need fewer licenses for expensive software because controlling software dynamically allocates them among users. (Yes, you're right—you can't install Flight Simulator on your own desktop; your tools manager controls what software is installed on the system.)

For those of you who started computing early enough, this setup looks strangely familiar. Long, long ago, if you had online access to computing, it was by means of a "dumb terminal" connected by a dedicated line to something called a "mainframe." The Sun executive, understandably perhaps, got a little upset that I made this comparison, and it's true that this new incarnation includes features that no one had yet dreamt of in the mainframe days. Load-sharing software is perhaps the key element, allowing any mix of jobs—from many small jobs running on one CPU to one big job shared across many CPUs. Also, the use of Smart-Card IDs in the network-access boxes allow you to "hot-desk" ("hot-screen") and instantly restore your own workspace, as you last left it, on any machine on the network you choose to sit at. All the same, I can't help but remark on the way the circle has closed, with the tide of PC evolution sweeping workstation-computer power out of the computer room and now once more concentrating that power as a central resource.

Consider, also, the latest oscilloscope range from Agilent (see "Lost that analogue-scope feeling?"). When analogue scopes first gave way to digital storage scopes and the signal you were looking at no longer directly drove the electron beam that painted the waveform on the display, it probably seemed obvious that the first part you could dispense with was the CRT, with its mass of glass and its uncomfortably high voltage demands. But the LCDs that followed just didn't look the same and didn't give you the same confidence that what you saw on screen truly reflected the signal you had tapped into. All of the major scope makers faced the same problem and applied immense ingenuity to digital processing algorithms and "simulated persistence" in an attempt to reproduce the preferred look and feel of the old scopes. Now Agilent has come up with an answer: Because everyone likes CRTs so much, fit a CRT—not the old analogue-scope CRT, of course, with its long tube and complex electrode geometry—but a modern graphics device that displays a heavily processed digital image. The fundamental appearance includes characteristics that we all prefer.

I do not claim that you can draw any deep message from this reappearance of venerable methodologies; new products have superseded many old products, and many of the old products are gone forever. But ideas that worked in the past did so for a reason. Even if technology has moved on, there's value in occasionally glancing backward to see if there's anything there to adopt and adapt. It's just one part of the much-promoted "design reuse," but this part comes for free.

Author info