| Summary | Very high load when 100+ concurrent users |
| Queue | Horde Groupware Webmail Edition |
| Queue Version | 1.0.3 |
| Type | Enhancement |
| State | Rejected |
| Priority | 2. Medium |
| Owners | |
| Requester | lumir.jasiok (at) vsb (dot) cz |
| Created | 11/07/2007 (6484 days ago) |
| Due | |
| Updated | 11/24/2007 (6467 days ago) |
| Assigned | |
| Resolved | 11/24/2007 (6467 days ago) |
| Milestone | |
| Patch | No |
State ⇒ Feedback
anyone but postgres users because of the regex match.
That regex is a bit confusing to me - what are you checking for
exactly? Why not use a subselect if it's supported (assuming we can
tell - don't worry about that for now) and otherwise the other query?
And can you please take a look at:
http://cvs.horde.org/framework/DataTree/DataTree/sql.php
to see the latest SQL, and perhaps update your patch for it?
Thanks!
that we already improved the DataTree perfomance in HEAD in a more
portable way.
this little patch improve performance in case that you have SQL which
have inner SELECT supported? I think, that this could help a lot. We
had load at server 25+ with 100 concurrent users (because of one wrong
SQL command!)and now we have load under 0.7 with 150 concurrent users.
That's a difference.
that we already improved the DataTree perfomance in HEAD in a more
portable way.
SQLite have inner select. MySQL < 4.1 is no longer supported afaik.
State ⇒ Rejected
Priority ⇒ 2. Medium
Priority ⇒ 3. High
State ⇒ New
New Attachment: leftjoin_to_innerselect-hwe-1.0.3.patch
Queue ⇒ Horde Groupware Webmail Edition
Due ⇒ 01/01/2008
Summary ⇒ Very high load when 100+ concurrent users
Type ⇒ Enhancement
slow - even on better HW and OS! That's why I did a research of the
code and optimized one of key SQL queris so that it now works ~40
times faster with PostgreSQL. It's been tested by thousands of our
users in production environment of my University for two weeks without
negative feedback.
The point of the enclosed patch is to replace a LEFT JOIN with inner
SELECT. I was wondering if you made it that way because of some
compatibility issues?
I am aware that the patch may not be considered up to your coding
standards, but for me it works. If possible, please merge the code as
soon as possible so that next security patches won't break the hack
for me :-)
Sincerely,
Lumir Jasiok, Ivan Dolezal