This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.

Bug 133115 - Time to create sequence diagram from 30 nodes exceeds threshold
Summary: Time to create sequence diagram from 30 nodes exceeds threshold
Status: RESOLVED FIXED
Alias: None
Product: uml
Classification: Unclassified
Component: Diagram Sequence (show other bugs)
Version: 6.x
Hardware: All All
: P2 blocker (vote)
Assignee: Sergey Petrov
URL: http://xtest-db.czech/AllInOnePerform...
Keywords: PERFORMANCE, REGRESSION, TEST
Depends on:
Blocks:
 
Reported: 2008-04-17 15:37 UTC by Alexander Kouznetsov
Modified: 2008-10-12 15:32 UTC (History)
1 user (show)

See Also:
Issue Type: DEFECT
Exception Reporter:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Alexander Kouznetsov 2008-04-17 15:37:55 UTC
Product Version         = NetBeans IDE Dev (Build 200804161203)
  Operating System        = Windows XP version 5.1 running on x86
  Java; VM; Vendor        = 1.5.0_12; Java HotSpot(TM) Client VM 1.5.0_12-b04; Sun Microsystems Inc.

Tests report that time to create sequence diagram from 30 nodes exceeds threshold on recommended configuration. The result
is as following: "3 of the measuredTime(s) [ 6818 5783 5779 5743 ] > expectedTime[5000] - performance issue (it's ok if
the first usage is in boundary of 0 to 2*expectedTime)"

See also URL.

See also previous issue 108269
Comment 2 Alexander Kouznetsov 2008-06-10 14:36:56 UTC
Build number:    200806071203 

Tests indicate there is a huge regression in 1st usage time (23 sec vs. 8 sec), and in subsequent usage (20 sec vs. 7
sec). The threshold is 10 sec for 1st usage and 5 sec for subsequent usages. See also URL.
Comment 3 Alexander Kouznetsov 2008-06-10 15:58:45 UTC
Please disregard the last comment, performance data is not valid.
Comment 4 Peter Lam 2008-06-17 01:17:56 UTC
marking it fix so that it will be reverified in NB 6.5 builds.
Comment 5 Alexander Kouznetsov 2008-06-26 13:40:04 UTC
Tests and manual testing indicate that there is a huge regression in 1st usage time (22-36 sec vs. 8 sec), and in
subsequent usage (20 sec vs. 7 sec). The threshold is 10 sec for 1st usage and 5 sec for subsequent usages. See also URL. 

Regression since new UML implementation.
Comment 6 Sergey Petrov 2008-06-26 14:06:13 UTC
in order to be able to compare with my system I need test system configuration. otherwise it's hard to compare.
on my system I have regression about 4-5 seconds in 6.5 against 3 in 6.1 for consiquent usage
Comment 7 Sergey Petrov 2008-06-26 14:06:55 UTC
is it creation of sqd from 30 class nodes?
Comment 8 Alexander Kouznetsov 2008-06-26 14:16:24 UTC
Tests are run on recommended configuration. 30 classes from jEdit-model are used.
solaris 1GHz, RAM 1GBB 
windows vista Intel P4 CPU 2,66GHz, RAM 1GB 
windows xp Intel P4 CPU 2,66GHz, RAM 1GB  
Comment 9 Sergey Petrov 2008-06-26 15:23:20 UTC
and what platform have 30 seconds result?
Comment 10 Alexander Kouznetsov 2008-06-26 15:45:09 UTC
Check out URL field: Window Vista has the worst results.
Comment 11 Sergey Petrov 2008-06-26 15:52:42 UTC
url do not contain info, need to connect to swan to check, thanks
one more question, how fast was 6.1 on vista platform?
Comment 12 Alexander Kouznetsov 2008-06-26 16:04:43 UTC
Sorry, I incorrectly indicated Windows Vista as the one with the worst results. That's wrong, the worst results are on
Solaris 10. Builds prior to 20080504 had 4 sec for 1st usage and 3 sec for subsequent usages on Solaris 10.
Comment 13 Sergey Petrov 2008-08-29 17:37:55 UTC
http://hg.netbeans.org/uml-main/rev/eca6f2992cc9
http://hg.netbeans.org/uml-main/rev/5464d620af64

I have about 8 times improvement locally.
Comment 14 Trey Spiva 2008-08-29 18:08:49 UTC
Reviewed
Comment 15 Peter Lam 2008-09-02 23:03:46 UTC
mrkam, could you please reverified the recent fix on your test systems. Thanks.
Comment 16 Alexander Kouznetsov 2008-10-12 15:32:56 UTC
UML tests are switch off since UML is not part of NetBeans