Web-Server Broadcast Storm

BYTE has received a flood of e-mail criticizing our March Lab Report, "6 Servers Tangle on the Web." Here's a sampling of the comments.

"...picked the overpriced DEC server" even though

"...Digital fails one of their tests working with WAIS."

"...don't let us see what they're using as test data."

"Not real-world"..."Should have used a T1 line instead of a 10-Mbps link."

"[BYTE] snubbed Apple,"..."ignored Open Transport 1.1," and generally

used incorrect terminology to describe the Apple server.

First, our stated goal was to see which Web servers were best for high-volume use. Our test-bed scaled up to simulate nearly 10,000 simultaneous users. We consider price in our reviews, but price does not make or break a product's score. We gave the Di gital Equipment AlphaServer 1000 4/266 the highest marks because it not only performs well but also has considerable upward scalability in the storage, memory, and open slots that would be needed for some high-volume Web use.

Not your cup of tea? We made it very clear in the review that the Silicon Graphics WebForce Indy "offers an outstanding price/performance ratio" for busy, but lower-volume, sites. That might have been clearer still had we had space to include all the performance-versus-load tests that NSTL performed. We will be updating this review on our Web site ( http://www.byte.com ) sometime in June to include performance-versus-load data. In the future, we will selectively expand such unpublished data on our Web site.

High volume was also the key to our decision to use a 10-Mbps network link. Whi le T1 is used by many Internet servers, it's too light a link to simulate intranet use or high-volume public access. All the servers tested would have pegged the T1 with no problem.

Some features for Apple's Web server were awkwardly or inaccurately described. For example, the network connection was described as AUI, not Apple's AAUI. Most of the information in question was given to us by a new employee at Apple; in the future, we will extend our fact-checking process to better verify vendor-supplied specs.

Do we snub Apple? In the February issue, we gave the Power Mac 7500 and 8500 1995 Editors' Choice Awards. We didn't "ignore" Open Transport 1.1; it was not a shipping product at press time. We covered it in some depth in "How Copland Communicates" (April).

We stand by our choice of the Digital AlphaServer 1000 4/266 as the overall winner in this Lab Report, with SGI's WebForce Indy and Intergraph's TD-40 as able, if somewhat less scalable, runners-up. We recognize that other Web-site scenarios exist, and we will continue to review Web servers aimed at various segments of the market without preconceived notions. In short, we will maintain our long-standing cross-platform commitment.--Eds.