John Reiser wrote:
Maybe it's really just a lot of users behind the same gateway that just
found out the disk were available. But isn't that sort of thing why
bittorrent was invented?
I did try torrents. The demand was miniscule (4 or 2 clients total in several
days); the tracker gave up [perhaps because sharing was miniscule?]
Then I tried direct http:. The demand was larger, but still small, for a week
or more. I was unprepared for the response due to the [unknown to me]
"advertising" in the Fedora Project FAQ. I'm trying to resurrect the torrents.
I'm not a torrent expert but I thought that if the tracker also seeds
the worst case single-connection scenario is no worse than an http or
ftp download.
Looking at the Apache logs, that spate of 500 requests had a few original
requests (http status code 200) for each disc, then an average 100 or so "resume"
requests (http status code 206) for each disc, with approximately non-skewed
distribution of resume points throughout each file. Some resume requests
for the same disc were within one or two seconds of each other, at not-near
offsets. That supports your hypothesis of multiple users behind a gateway.
But why would there be a _hundred_ resume requests per file (an average of
20 to 25 or so per supposed user, per file)? Meanwhile, how many bytes
were transfered, versus requested in theory but never actually sent?
Maybe there is some bandwidth throttle on the gateway - or a filter
blocking the content that keeps breaking the connection and the clients
don't know about it so they keep retrying.
--
Les Mikesell
lesmikesell@xxxxxxxxx
