The patch records raw measurements in the
access.log and store.log files. We have software
that extracts statistics from the logs. Below are the results of this analysis
for participating proxies. Detailed "profiles" on participating
proxies are available.
Study:
Leaf Proxies
-
- Utrecht University Proxy in the Netherlands
- Monday, September 22, 1997
- Tromsø University Proxy in Norway
- Sunday, October 12, 1997
- Monday, October 13, 1997
- Australian Defence Force Academy Proxy
- Wednesday, October 29, 1997 (sync-kludge-present)
- Wednesday, November 5, 1997 (sync-kludge-free)
- Proxy of University of Veszprem in Hungary
- Thursday, October 30, 1997
- Friday, October 31, 1997
- Computer Research and Automation Institute in Hungary
- Wednesday, October 29, 1997 (last day as a child of iif)
- Thursday, October 30, 1997 (first day as a stand-alone proxy)
- Friday, October 31, 1997
Intermediate Proxies
-
Top Level Proxies
-
- SURFNet Proxy in the Netherlands
- Tuesday, September 30, 1997
- UNINETT Proxy in Norway
- Sunday, October 12, 1997
- Monday, October 13, 1997
- Hungarian Academic Network Proxy
- Thursday, October 30, 1997
- Friday, October 31, 1997
- Thursday, November 13, 1997 (double-read-bug-free)
- SV Proxy (NLANR) in the USA
- Sunday, October 12, 1997 (October 12-16 are "special": see comments for each day)
- Monday, October 13, 1997
- Tuesday, October 14, 1997
- Wednesday, October 15, 1997
- Thursday, October 16, 1997
- Friday, October 17, 1997
- Saturday, October 18, 1997 (with double-read-bug planted)
- Sunday, October 19, 1997 (with double-read-bug planted)
- Monday, October 20, 1997 (with double-read-bug planted)
-
- Friday, October 24, 1997
- Saturday, October 25, 1997
- Sunday, October 26, 1997
- Monday, October 27, 1997
- Tuesday, October 28, 1997
Compare:
Leaf vs. Leaf
-
Intermediate vs. Others
-
Leaf vs. Top Level
-
Top Level vs. Top Level
-
Weekend vs. Weekday
-
Stand-alone vs. Child
-
Double-read-bug-free vs. Double-read-bug-present
- We are still working on this one...
- sv (Friday vs. Saturday; different load, not fair)
- sv (Friday vs. Monday; different load, still not fair)
- iif
Sync-kludge-present vs. Sync-kludge-free
- We are still working on this one...
- adfa (note minor changes in swap delays)
Track Patterns:
Comments
Most graphs on these page describe the behavior of a proxy server during
the day. The x axis represents time in hours, UCT. Most
measurements were done using 20 minute slots. Currently we
analyze HTTP traffic only. ICP entries are ignored.
Morning and evening "peaks" on small proxies that are due to
small number of requests (hence, no statistical significance) should be
ignored. Usually you may click on a graph to see the details.