The following charts show the number of accesses to the email channel and the number of accesses to the dynamic page that the email channel points to when we're in the paranoid mode.
The results were surprizing to me, too!!
Carver01 had problems at about midnight on the 21st so MyUW quit
talking to it. That's about the time the static page got turned on.
Carver02 has more dynamic requests that static requests.
Wait a minute, so does Carver03.
Ahh, that explains it, Carver04 must have taken on more of the MyUW
load than the rest of the carvers (note that the scale is larger
here).
Well, there you have it. The value of 'Y' is slightly larger than the
value of 'X'!!
Go figure!!
We did manage to reduce the number of calls during the
6am - 7am spike, though.
MyUW doesn't have a very good balancing ability, but that's okay as
it frees the rest of the carvers to do real work. Notice that
Carver04 doesn't get used at all until the 21st and then nearly all
the load switches over to that server.
The load based DNS distribution does a whole lot better job of
distributing the load among the individual Carvers.