DBIx::SearchBuilder 1.15 + RT 3.2.3rc1

RT Users
1

RT Users–

Is anyone running the DBIx::SearchBuilder 1.15 + RT 3.2.3rc1 combo yet?

Some of you may remember my posts a few months back about a “trouble ticket”
(bad pun) in our RT, with a large number of small text entries that took 30s+
to load, every single time.

Last week I gave DBIx::SearchBuilder 1.15 a try (CPAN had it listed as
experimental at the time) and saw the 30s+ number, literally, get cut in half.
Runs of this same ticket were coming in at 14s-15s.

Tonight I noticed RT 3.2.3rc1 up on Index of /pub/rt/devel,
and gave it a go. I’m now see that same ticket, I kid you not, at 7.5s-7.9s
every single time.

Does anyone else have an hour or two sometime this weekend to try the same
experiment, and see what kind of results they get?

Jesse–

Are the above gains (both DBIx::SearchBuilder 1.15 and RT 3.2.3rc1) this
significant expected, or do you think there are some outside factors here? To
my knowledge, no other factors have been changed.

-jd

2

Last week I gave DBIx::SearchBuilder 1.15 a try (CPAN had it listed as
experimental at the time) and saw the 30s+ number, literally, get cut in
half. Runs of this same ticket were coming in at 14s-15s.

Tonight I noticed RT 3.2.3rc1 up on
Index of /pub/rt/devel, and gave it a go. I’m now see
that same ticket, I kid you not, at 7.5s-7.9s every single time.

We’ve been working quite hard to improve performance. We think that when
you’ll try 3.4, you’ll be even more pleasantly surprised. Things are so
much better that RT now defaults to recording page render times in the
lower right-hand corner of each page (in a nice light grey, in a div tag
that you can make disappear if you don’t like it).

Jesse–

Are the above gains (both DBIx::SearchBuilder 1.15 and RT 3.2.3rc1) this
significant expected, or do you think there are some outside factors here?
To my knowledge, no other factors have been changed.

A 4x increase with just 3.2.3 and SB 1.15 is surprsingly good, but not
completely outside the realm of possibility. Not everyone will such a
dramatic performance boost, but everybody should see improvement.

Best,
Jesse
3

We’ve been working quite hard to improve performance. We think that when
you’ll try 3.4, you’ll be even more pleasantly surprised. Things are so
much better that RT now defaults to recording page render times in the
lower right-hand corner of each page (in a nice light grey, in a div tag
that you can make disappear if you don’t like it).

I have toyed with RT 3.3 in another Apache instance, but probably not since RT
3.3.7 or so. Is the latest build on Index of /pub/rt/devel
is getting close to becoming RT 3.4.0rc1?

For the sake of rt-users, here’s the kind of platform we’re on:

Software:

  • sorta-Redhat 9.0 (most everything is hand compled, so it’s not exactly
    Redhat as you know it)
  • mysql 4.0.22 with InnoDB Tables
  • Apache 1.3.33
  • mod_perl1 1.29
  • perl 5.8.5
  • DBIx::SearchBuilder 1.15
  • RT 3.2.3rc1

Hardware-wise:

  • Pentium4 3GHz
  • 1.5GB DRAM
  • 2xmirrored IDE drives

-jd

4

Jason A. Diegmueller wrote:

We’ve been working quite hard to improve performance. We think that when
you’ll try 3.4, you’ll be even more pleasantly surprised. Things are so
much better that RT now defaults to recording page render times in the
lower right-hand corner of each page (in a nice light grey, in a div tag
that you can make disappear if you don’t like it).

I have toyed with RT 3.3 in another Apache instance, but probably not
since RT 3.3.7 or so. Is the latest build on
Index of /pub/rt/devel is getting close to becoming RT
3.4.0rc1?
It would closer if you install it on your another apache instance, make
snapshot from your real DB and play with it, test it for regressions.
Especially custom fields and all new interfaces. Report back on errors
or success. More testing is a required.

				Best regards. Ruslan.