In these sober post-dotcom crash days, we’re even more aware that the success of a web site is not based on the sheer quantity of hits but also their quality. Fortunately, all web server software collects traffic logs. We examined three applications that analyze these server logs to pull information from the long strings of numbers and acronyms: Flowerfire’s Sawmill 6.2.5, Summary.net’s Summary Pro 2.0.1, and the well-regarded, open source Analog 5.0.3. We found that while free will always be an enticing price, Summary Pro is worth the investment.
Slicing and Dicing
Before you can find out that the majority of traffic to your site comes from Canadian universities between 2:00 and 3:00 AM, your web analyzer has to sort through your log collecting and categorizing information. When you launch Sawmill, it opens up a tiny window with a single button that launches your web browser and opens the first Sawmill configuration page. You answer a few basic questions, such as where your log data is stored, and it then works its way through the data.
Summary’s startup behavior is a bit overeager. The first time you launch Summary it immediately begins attacking any log files in its Logs folder. We think that the first time you start up the program it should take you to the Summary configuration page.
Configuring Analog requires you to edit a text-based configuration file. Although this is not uncommon for software with Unix roots, it does leave you open to fouling up the instructions up with a typo. Sig Software’s Analog Helper (www.sigsoftware.com), whose latest version 1.5 offers full support for Analog 5’s new options, solves this problem by providing a graphical front end to Analog’s settings, generating correct config files, and then launching Analog.
Always in a Hurry
Speed is crucial to a web log analyzer. On some very busy sites web log analyzers are slower than the traffic hitting every second. On our test machine, a G4/450 with 448MB of RAM running Mac OS X 10.1, Sawmill processed about 3,600 lines per second with its default settings. Summary crunched through the same logs at around 20,000 lines per second, five times faster than Sawmill. Analog was even faster, crunching approximately 28,000 lines per second.
All three of these applications run in both Mac OS X and Mac OS 9. In our testing, we found the best performance when the applications were running on OS X and we were browsing the data using Internet Explorer on the same machine. Mac OS 9’s less impressive performance is almost certainly due to the way it handles process scheduling between a client and a server on the same machine. OS X offers a noticeable boost in responsiveness over Mac OS 9.
Filtering out the Noise
Once the initial crunch is done it’s time to fine tune the reports. For example, you may want to filter out the web traffic from within your own organization. Application speed is particularly crucial here as you may add and change filters many times before the data answers the questions you care about. All three analyzers provide ample filtering capabilities. Summary’s and Analog’s analysis speed is a major benefits here, though most web administrators will only need to perfect these settings once per web site.
Summary and Sawmill don’t simply crunch through the logs and spit out HTML files. Instead, they put the log data into a database you access through a web browser. Because both of these products are application-specific web servers in their own right, you view the data from any platform with a web browser. Analog only generates static HTML files as output.
Crunching alone isn’t enough, though: good analysis tools must create intelligible and useful reports. Sawmill gives you an overview of the results and a left-side navigation bar full of sub-reports. Although we found the overall quality of Sawmill’s graphs to be higher, Summary’s default report set is more evolved; for example, the sub-report that points out possible server problems is a huge time-saver for administrators. That said, all three tools generate the standard complement of traffic, bandwidth, and popularity reports.
Both Summary and Sawmill come in multiple versions to suite your log crunching needs; both are also available in time-limited demo versions. Sawmill’s basic package only permits a single user but will analyze multiple domains; pricing is tiered by increasing numbers of users. Summary Pro’s Standard version provides basic processing for up to three domains. Summary also has a Plus version which supports up to domains and adds incremental processing. Their SP version is aimed at service providers with up to 1,000 domains. The SP version also allows you to edit the HTML report templates by adding a logo or changing a header.
Macworld Buying Advice
For those on very tight budget, the combination of Analog and Sig Software’s $25 Analog Helper is workable. Analog’s blazing speed almost compensates for the lack of interface and very rudimentary graphics. Analog without Analog Helper is only for the truly stout of heart; getting your analog.cfg file just right is odious without it.
While both Summary and Sawmill are able to turn raw data into comprehensible information, Summary Pro’s speed and pricing give it an edge. We do wish that some of the SP features were available in the Plus version, especially the HTML customization capability.