Nerd Alert: Switching Web Hosting Companies

In an earlier post I mentioned I took a major step this past month in changing Web hosting companies. I had been a very satisfied customer of PowWeb for 6 years. However, during the last 3 months my service with them began to degrade. After several failed attempts working with their technical support to get things back where I felt they should be, I decided to make the switch to Lunarpages. So far, I have been very, very pleased.

In terms of price and features, both PowWeb and Lunarpages are essentially equal. They are both about $7 per month for all the Web and data services one needs to run a site like www.lesia.com. Lunarpages outshines PowWeb significantly when it comes to performance.

I had been hosting a couple domains including lesia.com with PowWeb. Beginning around January, I noticed the response time (how long it takes the Web site to respond when clicking through the site) was getting longer and longer and the transfer rate (how quickly pages and images are brought down to the browser once the site responds) was getting slower and slower. Of course, I was still being charged the same amount despite this degraded service. The only way to know for sure was to run some simple tests to compare PowWeb’s response time, route mapping and transfer rate with other popular and similarly priced offerings.

To find a vetted list of similarly priced and comparable offerings, I consulted the Top 10 Hosting Reviews site. As you can see price, disk space and bandwidth vary significantly among the sites on the list but none of those variables were particularly important to me. I wanted speed. Next came the tests.

Response Time

To test response time I used the Ping command-line program that has been included with Windows (and every other operating system) since local area networks became popular. Ping is used to determine whether another computer is reachable over the network. Similar to sonar ping used by submarines, the utility sends a specially-crafted request to the remote computer in question and waits for a response. If the remote computer is active and configured correctly, it will see the request and reply with a specially-crafted response. Based on how long it takes the remote computer’s response to be returned we can estimate the roundtrip response time.

Below is a screenshot of the ping test I conducted against a PowWeb server (first 3 red boxes) then a Lunarpages server (second 3 red boxes).

PowWeb vs. Lunarpages ping test

Notice the “time=” entries for PowWeb are nearly twice those for Lunarpages. The averages are 108ms vs. 59ms, again, roughly double. That means it takes twice as long, on average, for my computer to get a response from the PowWeb hosting server than it takes for the Lunarpages server. Sure, we are only talking milliseconds here (don’t blink) but this delay occurs for each and every request (ignore caching for now) including all those pretty pictures, emoticons, scripts, formatting rules, etc. Just visiting the homepage of www.lesia.com results in about 20 requests back to the lesia.com Web server. Even with request pipelining the 2x difference is noticeable.

Response Time Winner: Lunarpages

Route Mapping

To test route mapping I used the Traceroute command-line program that also has an equivalent included with every modern operating system. Traceroute is used to determine the path and time a typical request takes to go from one machine on the network to the other. Telling your browser to go to www.lesia.com never results in a direct request between your machine and the server hosting that site. There are firewalls, gateways, proxies, caching servers, load balancers, and a bunch of other types of machines your requests must traverse before reaching the server(s) actually hosting the site. The resulting responses to those requests follow a similar path back before anything gets displayed in your browser. What does this mean? Well the more machines your request must traverse (these are called “hops”) the longer (on average) the delay. Each of those hops incur a separate response time penalty. Remember, the shortest distance between two points is a straight line. The more hops, the more left and right turns your request takes en route to its destination.

Below is a screenshot of the traceroute test I conducted against a PowWeb server (first 2 red boxes) then a Lunarpages server (second 2 red boxes).

PowWeb vs. Lunarpages traceroute test

In this test you can see there are 20 hops between my machine and PowWeb and 14 hops between my machine and Lunarpages. Lunarpages requests take 30% fewer left & right turns than the same requests to PowWeb from my machine. Now this would not be a big deal if the latency (response time) along the PowWeb hops was the same or lower than the Lunarpages hops. To summarize, there are more hops between my machine and PowWeb than Lunarpages and those PowWeb hops take longer. We knew the trip (sum of the hops) takes longer from the above Response Time test as well.

Route Mapping Winner: Lunarpages

Transfer Rate

It is difficult to conduct a transfer rate test without having control over the files on each server being compared. Because of caching, network congestion and other factors beyond my direct control, it is necessary to conduct remote transfer rate tests during different times and for different file sizes. To do that, I actually needed to take the leap and purchase a Lunarpages account. I was fairly sure the Lunarpages transfer rate would not be any worse than PowWeb after surfing various sites hosted by each. Also, I was fairly sure I would like Lunarpages more than PowWeb after the Response Time and Route Mapping tests. So, I whipped out the plastico fantastico and signed up. (By the way, Lunarpages offers a 90-day money back guarantee when transferring from another hosting company.)

I conducted a series of real-world transfer rate tests including downloading videos, viewing photo galleries, transferring content via FTP, importing/exporting databases, etc. I performed these tests over the course of a week including the weekend and at various times of the day. Lunarpages was consistently significantly faster than PowWeb for both downloading and uploading.

Transfer Rate Winner: Lunarpages

Other

When comparing services we sometimes overlook some of the non-tangible or less measurable aspects of performance. Getting things done is not only about quickly shuttling bits and bytes from one machine to the other. Learning how to shuttle those bits and having the proper tools to assist in shuttling those bytes are equally important. Accordingly, I included a couple additional soft metrics in my evaluation: Control Panel interface and User Forums quality.

All big hosting companies have browser-based site administration interfaces. These control panels provide the knobs, levers and buttons a site operator like myself would use to configure things to our liking. There are good ones. There are bad ones. I have yet to see a great one. From my experience, the PowWeb control panel beats Lunarpages hands down. The PowWeb control panel is better organized and better designed making it is easier to find what I need.

PowWeb Control Panel Lunarpages Control Panel

Control Panel Winner: PowWeb

The profit margins in the hosting business are razor thin these days. As a result, a lot of hosting companies maintain user forums where customers can help each other troubleshoot, diagnose and fix problems they encounter while running their sites. It is usually impossible to get help over the telephone and chances are, someone else has had the same issue you are having. Getting customers to shoulder some of the technical support and documentation burden keeps many of these companies in business. The hosting companies also use these forums to communicate maintenance schedules, upgrade plans, and other information quickly and efficiently to their tens- and hundreds-of-thousands site operators.

Both PowWeb and Lunapages operate well-designed, well-organized user forums which both have active community participation.

User Forums Winner: Draw

Summary

Here is a table that captures the results of this head-to-head comparison:

PowWeb vs. Lunarpages summary table

Lunarpages wins this match-up. The only question that remains is whether they can stand the test of time as PowWeb did for 6 years. Changing hosting companies is a ton of work and not something I want to do very often. Once you find a good one there is often very little incentive to switch. In my case, the performance issues pushed me over the edge and I am pleased to have found a new hosting company that meets my needs for the same price.