Re: CVS performance affected by tags?

[Eric Siegerman]

Another question would be whether you have long-lived,
still-active branches.  Branches slow down much more quickly as
their length increases than the trunk does.  In fact, a branch
also slows down as the number of *trunk* revisions since the
branch point increases.

This is because of the way revisions are stored in the repo.
The latest trunk revision is stored intact; the previous
trunk revision is stored as a reverse delta from the latest,
and so on.  Branches are stored as forward deltas from the branch
point.

   1.1<--1.2<--1.3<--1.4
          \
           \
            >1.2.2.1-->1.2.2.2-->1.2.2.3

In this file, the only revision stored intact is 1.4.  To get any
other revision, CVS internally applies patch after patch after
patch, as shown by the direction of the arrows.

Committing 1.2.2.4 is straightforward: compute the differences to
get from 1.2.2.3 to 1.2.2.4, and store them.  But committing 1.5
is a bit bent:
  - store 1.5 as-is
  - compute differences to get *from* 1.5 *to* 1.4
  - replace rev. 1.4 in the repo with those differences

As you can see, committing either 1.2.2.4 *or* 1.5 will add
another patch that needs to be applied to recover the latest
branch revision.


On Wed, Aug 22, 2001 at 02:52:44PM -0400, Donald Sharp wrote:
> What is happening on the cvs server? 
> 
> What does the load and io wait time look like?
> Does the machine have enough memory?( Is it swapping? ).
> 
> In general though the more versions and tags that 
> you have in a repository, the slower cvs will be.
> 
> donald
> 
> On Wed, Aug 22, 2001 at 11:25:52AM -0700, Paul D. Schmidt wrote:
> > Our cvs server has been slowly degrading in performance
> > (at least with checkouts and updates)... would removing
> > some of our ~850 tags help that at all, or should I be
> > looking at something else?
> > 
> > Thanks,
> > -Paul
> > 
> > --
> > [Paul D. address@hidden
> > [Xtime, Inc................][........http://www.xtime.com]
> > 
> > 
> > _______________________________________________
> > Info-cvs mailing list
> > address@hidden
> > http://mail.gnu.org/mailman/listinfo/info-cvs
> 
> _______________________________________________
> Info-cvs mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/info-cvs
--

|  | /\
|-_|/  >   Eric Siegerman, Toronto, Ont.        address@hidden
|  |  /
With sufficient thrust, pigs fly just fine. However, this is not
necessarily a good idea.
        - RFC 1925 (quoting an unnamed source)