[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Thu, 14 Dec 2000 13:14:00 -0800
Performance can be addressed several ways.
If you have a somewhat large module, but you only need a subset of it,
consider either breaking the module up into smaller modules or creating
entries in the modules file such that when you perform a "cvs get module"
you are only receiving what you need. The caveat is that if you are using a
"module" defined in the modules file as opposed to a module under $CVSROOT,
then you should use "cvs update -d" only where necessary to pickup new
directories. Otherwise you will be picking up all of the directories that
the original "cvs get" didn't picked up whether you wanted them or not,
thereby negating the reason you setup the modules file entry.
Bottlenecks to the CVS server are usually, in order of severity, network,
disk I/O and finally CPU (sometimes memory). Having a good admin that can
monitor what is happening and make the appropriate tuning tweaks and
recommendations is invaluable.
Methods of increasing network throughput for your specific needs can differ
for almost every case. It is important to understand where the network
bottlenecks exist. I have done several things over the years to improve
network performance, such as re-segmenting the network, replacing hubs with
switches, replacing 10Base-T equipment with 100Base-T, using dedicated IP
addresses for specific applications, purchasing faster dedicated lines
between sites, dedicating a minimum percentage of network bandwidth for a
specific application. However, in each case it was specific to the
situation. There is rarely a one size fits all solution and cost can play a
major factor. I would suggest talking to the appropriate person in IT/MIS
to see what can be done.
There are a whole host of disk I/O solutions as well. For the most part,
most of the disk intensive servers I setup, I use hardware RAID setup with
RAID 0+1 (striping over separate SCSI channels), using the fastest drives
and fastest I/O solution I can get. If users will tend to be reading and
writing many of the same files and directories, you might look into storage
solutions that provide battery backed cache (not cheap).
If your CPU or memory bound you may want to try tuning the system and/or
Hope this helped.
From: James Stauffer [mailto:address@hidden
Sent: Thursday, December 14, 2000 9:49 AM
What can be done to improve the performance of CVS? Is this topic
already covered on some website or FAQ. We use pserver to connect to
a remote RedHat machine.
Info-cvs mailing list