In Search of Speed

Home > Build > Design > Articles

by Bob McElwain

Among Webmasters, the search for faster page download speeds continues. The topic comes up frequently. And the usual suggestions are forthcoming. Minimize the use of graphics and optimize the size of those used. Use width and height with the image tags, for it allows text to load quickly, which gives the visitor something to jump into right away. And there are others, such as doing without Java script if possible.

However, the recommendation of finding a fast server continues to rank high on most lists. I have recently concluded this is not as important as some believe it to be. If you have a good host that is not overloaded, you probably will not improve your page download speeds significantly by moving to a "faster" server. Here's why.

One of my clients is a software developer who markets his products via the Web. The demands of Windows 95 and 98 have resulted in an enormous increase in the size of computer programs.

Prior to the introduction of Windows, significant MS-DOS programs were typically about 250K bytes when zipped. Now they are frequently over 5 megabytes.

Even with increased modem speeds this tries the patience of potential customers who elect to download the trial version, for it can take a very long time. And given an interrupt, more likely with the longer download times, one gets to do it all over again. Or give up, which to us means the loss of a potential customer.

In an effort to improve download speeds, and thus also decrease interrupts, I explored a number of options, including building our own server. I put a lot of time into this, and largely wasted all of it.

Had I had the thinking cap on straight initially, I could have put it all together in very short order. Hopefully what I discovered will assure you there is no need to experiment further, and likely no need to search for a faster host for your web site.

I have used Pair.com for years. Now I also use JumpLine.com Virtualis.com is highly recommended by many, so I opened an account with them as well. I tested many factors, including upload speeds via FTP. But I will limit the numbers here to file download speeds, for this is what relates directly to page download speeds.

After uploading a 3.3 megabyte file to each site, I checked download times from all three. Beginning on a Monday at 7 am New York time, I made the first download. When it completed, I started a download from the next site. When it finished, I went to the third. I continued the cycle throughout the day, wrapping at 6 PM New York time. I used an older slower machine, a 486 with a 33 KB modem. In all, I downloaded six times from each site. The best time was 21 minutes 38 seconds, the worst, 22 minutes 59 seconds. Both were from JumpLine. All other times were between these two.

I had planned to repeat the experiment throughout the week. But when I got virtually the same results the following day, I called it quits. It was clear that all three were equally fast.

It was not what I expected. I pay $5.95/month at Pair.com, $24.95/month at JumpLine and for that month, paid about $49 at Vitualis. I had expected better performance from one of the three. Looking at pricing, I was guessing it might be Vitualis. Further, they are close geographically to me, some 150 crow-fly miles to the south. That there was no significant difference surprised me, but it shouldn't have.

If you follow any ezine for a time, you will hear folks say your page should download in 5 seconds, or whatever they believe. Or somebody you trust will say your pages really download fast; only 8 seconds. Another says the same, but gives 15 seconds. Why are the numbers so different? How can one fellow get my pages in 8 seconds when it takes another 15, yet both believe it is fast?

The answer lies in our individual connections to the Web. While the modem, disk, and CPU speed of your system are factors, the route traveled to your ISP (Internet Service Provider) and its performance are also critical.

We all face such limitations, as do our visitors. We can do nothing about the speed with which our visitors access our site, for their system and ISP determine this, and perhaps to some extent, the paths available to them through the Web.

As an example of the latter, I do fine getting through my ISP, but their only route is through Fresno, California, and for some reason this is a bottleneck.

As Webmasters, all we can do is be sure we have a fast host so that we are not part of the problem. But before you run off in search of a better one, check with some care. If you have a good host, one that is not overloaded, it is doubtful you can do significantly better by moving.