The business of Web Hosting

The low-downs and downs and downs

I figured it would be a good day to get these ISP-related posts that have been waiting to be published out. In reading all the great questions and responses about web hosting here and recently in Apart, I decided to interject some of my limited knowledge on the subject. I have concentrated my efforts as of late on different areas other than web hosting, so please correct me if I present any outdated information.

I think the largest misconception right now is that price and quality are proportional. While this should be true, we do have an industry in its infancy and a lot of naive people wanting its services. Not that being naive is bad (you can't know everything) but it does present an opportunity for price gouging. This practice, at least for the time being, does not appear to exclude large and reputable companies.

Basically, a quality web hosting service will sustain a cost of $.06/MB of data transfer per month. This figure includes tech support, equipment, equipment upgrades, rent, etc., etc. Basically the whole shot. Everything in web hosting can be broken down into a bandwidth dependent cost due to the nature of the business. This is one of the main reasons to be wary of 'unlimited bandwidth' web hosting plans. I will explain more on this below. Of course the $.06/MB can be reduced in various ways including outdated equipment, overloaded connections, crappy support, etc. Offering improved reliability, supercomputers, etc can also increase this figure. These are features that I personally feel begin to get into the "excessive and unnecessary" arena... rather, there is no ROI.

When you look for web hosting in the open Internet market, it can get very competitive. When you simply go to your local ISP you can expect to pay through the nose (sometimes, see price gouging above). This competitiveness has resulted in some interesting and very unprofitable hosting plan offerings. But because the plan is 'unprofitable', does not mean the service in general is. A fairly common offering right now are plans with 2000MB of data transfer, 50MB (or so) of disk space. Well, if you use $.06/MB x 2000MB this plan should COST the ISP $120/month. Then of course they must make a profit above that.

So, does this mean to get such a plan you must pay $150+/month for such a hosting plan? NO, and this is where the competitiveness comes in. Since people tend to flock to the service provider that gives the most bang for the buck, service providers offer 'unprofitable' plans to get the customers. The gamble is that the average user will only use 300MB (or similar number) data transfer per month anyhow, so why not offer a lot of bandwidth?? The truth be known, if you USE 2000MB/month in data transfer and you pay < $120/month for a plan such as the one listed above, the ISP is actually LOSING money on you. It's a battle of averages and you'd be on the winning end.

The above scenario is a gamble and as bandwidth gets larger and websites get bigger, obviously the average website data throughput will increase. Expect prices to rise in the future, UNLESS bandwidth gets cheaper (what I am hoping for). This is already becoming a problem in the dial-up ISP market and we are noticing plan changes and/or increases in price (i.e. AOL and IBM). So what about unlimited bandwidth plans? Well, the cost/MB throughput does not go away and this is a bigger gamble than the example above. Examine service contracts closely and pay special attention to 'abuse' clauses. Many will consider excessive bandwidth use 'abuse' and suspend your account. Not really unlimited bandwidth after all. If you pay $30/month and utilize over 500MB thruput/month they are losing money on you or they are cutting corners. Be wary, and remember that *someone* needs to pay for the high bandwidth users.

So what makes a quality web hosting outfit?

  • Servers - Quality servers are important. Contrary to popular belief, the number of accounts on a machine does not reflect an ISP's quality. Some machines will accommodate 100 users where others might accommodate 1000 users. That's right, 1000 web accounts on a single machine is not uncommon, especially if that machine cost $30,000+.

  • System Administrators - you can have the best machines on the world, but if the dingbat behind the keyboard does not know how to set things up properly it will crash and crash often.

  • Redundancy - multiple connections is nice to have. If a line goes down another can pull the slack. Even better is multiple connections to multiple providers. That way if Sprint went down (for example) your connection to UUNet can handle things.

  • Support - support is always important, but different people do need different levels of support.

  • Network - The backbone of everything, if they don't use quality routers, etc. you can have a big bottleneck. This is probably the most underrated feature of a web hosting outfit, make no mistake that it IS one of the most important.

  • Backup - a quality provider will make daily backups of data and will have reliable backup power.

  • Connection - 50% average bandwidth utilization is a lot, look for less. T1? T3? OC3? DS3? (the meaning of these is another long message) The size of the pipe does not matter if it is not being fully utilized. A T3 is NOT necessarily better than a T1! (Its not that size that counts, its how you use it. :) The provider, hops to backbone and about 100 other things can also make a difference here. A full T1 will realize 1.54Mbps and a full T3 will be ~ 45Mbps (28 T1 lines, but you do loose some bandwidth with 'trunking'). Mbps = megabits per second, this is NOT SPEED. A packet of data travels at the same speed (i forget the number offhand) no matter what pipe it goes through. What is measured is the AMOUNT of data it can transfer at that speed.

There are hundreds of factors and I could go on for a while more, but the above are major ones to be familiar with.

How do you know if your ISP measures up?? You don't. They can throw big names at you like 'we have $100K servers and 10 T3 lines' but that does not mean they are good. That's what makes this so difficult. My best advice is to listen to your colleagues and go where they have had the best luck, but be aware of what you are looking for.