The Server Log
- The Web server log collects the following data for every Web page
- OK/not hit (e.g. file not found))
- source of hit (key)
- what file was hit (key)
- how many bytes were downloaded (session vs. page)
- We have though about aggregation and analysis of server
log data but have not really automated it yet. At this point
we have more questions than anything else.
- How long should we keep the detail? How long will UMS keep it
- What summary statistics should we calculate? Number of files? Number
- What class variable should be used?
- File, directory, or group of directories?
- IP or groups of IP addresses?
- How we assure comparable user statistics over time?
- Are three-stage Markov chains a good way for generating
a model of user access behavior?
- It seems like most of the tools that are out there are not nearly as
flexible as we'd like. We want MICS or SAS/CPE.