Re: Subversion trunk (r1078338) HTTP(/WC?) performance problems?

Hyrum K Wright <hyrum_at_hyrumwright.org> writes:

> Having only tangentially followed this thread, I've got a couple of questions:
> * Do the current subversion+serf fixes solve the the large checkout
> time problem people have been seeing on trunk?

I have a mirror of the Subversion repo (mod_dav_svn 1.6). Checking out
trunk over a LAN gives onto a local disk:

1.6/neon
real 0m8.661s 0m8.509s
user 0m3.710s 0m3.790s
sys 0m1.920s 0m2.020s

1.7/neon
real 0m8.814s 0m8.779s
user 0m4.070s 0m4.280s
sys 0m2.710s 0m2.510s

1.7/serf
real 0m8.186s 0m7.973s
user 0m3.850s 0m3.650s
sys 0m3.310s 0m3.410s

So 1.7 uses more CPU but is about the same elapsed time, if anything
1.7/serf is now faster than 1.6.

Network disks are still a concern. Checkout onto NFS (mounted sync):

1.6/neon
real 1m58.342s
user 0m5.990s
sys 0m10.830s

1.7/serf
real 7m13.181s
user 0m8.450s
sys 0m35.300s

Checkout onto NFS (mounted async):

1.6/neon
real 1m26.692s
user 0m6.150s
sys 0m10.690s

1.7/serf
real 6m20.883s
user 0m11.040s
sys 0m37.460s

We using a lot more CPU and elapsed time is hugely increased.

-- 
Philip