On Wed, 26 Jul 2000, Bjorn Townsend wrote:
> We recently tried an interesting experiment -- we opened pipes on
> our clients *and* our servers with ipfw, and noticed tremendously
> improved performance. However, in the polygraph documentation, the
> instructions say to only run "ipfw -f flush" on the clients, and
> only add pipes on the servers. Is this an error in documentation, or
> is there a reason why things are that way?
There is at least one reason. We are simulating a "corporate" network
environment (fast local clients, slow remote servers). We used to have
pipes on both sides for the "ISP" environment (modem clients, slow
remote servers), but the majority of users were less interested in that.
When you are talking about "tremendously improved performance", what do
you mean exactly? What was improved (response time, request rate, memory
usage, etc) and where (clients, DUT, servers, etc.)? Please share the
details if possible.
By installing pipes on the client side, you start shaping the outgoing
traffic so performance differences are quite possible.
Alex.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:14 MDT