Mac Web Server Performance
Monday, June 17, 1996 by Dave Winer.
The Mac OS was designed for individual users, and in that role it performs very well. But the performance of its file system appears to be a major performance limit for high traffic Mac-hosted web servers. I've mentioned this several times in past DaveNet releases.
In late May I decided to run some tests to see if this belief was true (it was) and then I started a collaboration with Chuck Shotton, firstname.lastname@example.org, the developer of the market-leading WebStar web server software, published by Quarterdeck; and Doug Baron, email@example.com, my co-developer on Frontier.
We're pretty far along, and hope to release the new technology, some from UserLand's source code base, and other parts using new connection technology in WebStar. It will fit in with both our architectures, WebStar's plug-in interface, and Frontier 4.0's website framework. I believe it will up the flow on high traffic sites by as much as a four-to-one ratio. Something to look forward to! I know I'm sticking my neck out, but I think we'll deliver.
Since both Chuck and I have invested heavily in the Mac system and its connection to the web, it made sense for us to connect our stuff together more intimately, to make our competitive position relative to other platforms stronger. So that's what we're doing. We have fun playing together, and I think that shows in our work.
Over the weekend, I posted the results of a more formal performance test. It shows part of my contribution to the collaboration. I wanted to include the DaveNet readership in this loop. When we have an announcement to make, which should be soon (maybe at MacHack?) having laid this groundwork will help me to explain what we've done.
The page is at http://www.scripting.com/lab/filesystem.html.
If you're interested in cross-platform web technology, please check it out.
I owe Chris Espinosa, firstname.lastname@example.org, a hearty thank you for changing my way of viewing the world. At lunch a couple of weeks ago he said that the Mac has 33 percent market share. I looked at him, puzzled, I wanted to know more!
He said that in the world of the Internet, the cross-platform story is the key. I agreed. Java says it's cross-platform. Netscape is cross-platform. Cross-platform is such a key criteria that even Microsoft has renewed their investment in the Macintosh, and is even considering investing in Unix software.
If you value your independence of Microsoft, acknowledge the Macintosh. And if you're Microsoft, acknowledge the Macintosh because everyone else does. I find that the inverse is true. I'm looking for parts of my code base that can be ported to Unix and Windows. Once I started looking seriously, without fear, it became clear that large pieces of my technology can be easily ported, and can easily make a difference on other platforms.
It's a three-way world. Some of us specialize in one of the ways, others cut across all the ways. There are advantages to both approaches, and in the evolving ecology of the Internet, specialization *will* matter. People who can deliver the power of a platform and also support cross-platform technology will co-exist. It'll be better if they share their ideas. Of course.
The Macintosh user base specializes in content. Windows makes a powerful, inexpensive client. Sysops think of Unix when they think of servers. Sure the markets slosh around, and eventually it isn't that simple anymore (a Mac makes a great server!) but there's room for three platforms, and actual number of users is a very unrefined way to shape a strategy. Client, content and server, those are the three markets. Windows, Macintosh and Unix, those are the platforms.
Right now Java owns the cursor. Its cross-platform potential is its biggest selling point, but also is unspecified. Guess where we'll be digging next? Cooool.
PS: Chuck's company, Biap Systems, is an acronym for Brain in a Pan. Hey, Chuck's a funny guy!
PPS: UserLand is not an acronym.