I wondered if some of the recent changes in Launchpad would have changed this. There is definitely still a difference:
mbp@joy% time bzr -Dhpss -Dhpssdetails push bzr+ssh://chinstrap.canonical.com//tmp/docdiff-test2
Created new branch.
HPSS calls: 14 (1 vfs) SmartSSHClientMedium(bzr+ssh://<email address hidden>/)
bzr -Dhpss -Dhpssdetails push 0.24s user 0.06s system 2% cpu 13.296 total
mbp@joy% time bzr -Dhpss -Dhpssdetails push lp:~mbp/+junk/docdiff-test2
Created new branch.
HPSS calls: 14 (1 vfs) SmartSSHClientMedium(bzr+ssh://<email address hidden>/)
bzr -Dhpss -Dhpssdetails push lp:~mbp/+junk/docdiff-test2 0.27s user 0.05s system 1% cpu 19.633 total
so 6.34s or nearly 50% slower. It's doing the same number of calls and the same amount of traffic; lp just has more lag.
The differences are:
* 2.4s extra delay before starting the SSH connection, I think because we're doing a separate xmlrpc to find the URL (bug 397739) (2.551s vs 0.168s)
* 1.1s extra delay opening the SSH connection and getting the first response back (6.7s vs 5.63s); this might be fixed by bug 660264 when that's live
* 2.5s extra delay in BzrDirFormat.initialize_ex_1.16 (3.0 vs 0.46), possibly because this is doing slow permission or other checks on the server side before proceeding
* 0.34s other extra delay extra
I wondered if some of the recent changes in Launchpad would have changed this. There is definitely still a difference:
mbp@joy% time bzr -Dhpss -Dhpssdetails push bzr+ssh: //chinstrap. canonical. com//tmp/ docdiff- test2 edium(bzr+ ssh://< email address hidden>/)
Created new branch.
HPSS calls: 14 (1 vfs) SmartSSHClientM
bzr -Dhpss -Dhpssdetails push 0.24s user 0.06s system 2% cpu 13.296 total
mbp@joy% time bzr -Dhpss -Dhpssdetails push lp:~mbp/+junk/docdiff-test2 edium(bzr+ ssh://< email address hidden>/)
Created new branch.
HPSS calls: 14 (1 vfs) SmartSSHClientM
bzr -Dhpss -Dhpssdetails push lp:~mbp/+junk/docdiff-test2 0.27s user 0.05s system 1% cpu 19.633 total
so 6.34s or nearly 50% slower. It's doing the same number of calls and the same amount of traffic; lp just has more lag.
The differences are:
* 2.4s extra delay before starting the SSH connection, I think because we're doing a separate xmlrpc to find the URL (bug 397739) (2.551s vs 0.168s) initialize_ ex_1.16 (3.0 vs 0.46), possibly because this is doing slow permission or other checks on the server side before proceeding
* 1.1s extra delay opening the SSH connection and getting the first response back (6.7s vs 5.63s); this might be fixed by bug 660264 when that's live
* 2.5s extra delay in BzrDirFormat.
* 0.34s other extra delay extra