You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@jmeter.apache.org by bu...@apache.org on 2019/08/06 04:20:48 UTC

[Bug 63639] New: give you three times qps improve

https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

            Bug ID: 63639
           Summary: give you three times qps improve
           Product: JMeter
           Version: 5.0
          Hardware: PC
            Status: NEW
          Severity: normal
          Priority: P2
         Component: HTTP
          Assignee: issues@jmeter.apache.org
          Reporter: 206763753@qq.com
  Target Milestone: JMETER_5.2

i write my jxm as the attachment image sax.png, with 31 or 100 threads ,i got
6300 qps
all.txt size is 113M, with 10800000 lines in.

then, i wtite 31 threads(Runable),igot 14000qps, many times.

so,i doubt that your file loader must have threads compete, and i suggest that
you can split the big test data file to thread number parts; so there is no
competition; i am sure this will improve the qps i can got

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

--- Comment #6 from Graham Russell <gr...@ham1.co.uk> ---
Which version of JMeter are you using? You said 5.0 in the meta data but 3.2.x
in your comment, or have I misunderstood?

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

--- Comment #5 from Felix Schumacher <fe...@internetallee.de> ---
Created attachment 36725
  --> https://bz.apache.org/bugzilla/attachment.cgi?id=36725&action=edit
Finer locks for FileServer

This is mainly a test, whether finer grained locks for FileEntries in
FileServer would speed up multi-threaded usage of files in CSV data source.

It didn't speed up things on my setup.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

Philippe Mouawad <p....@ubik-ingenierie.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |p.mouawad@ubik-ingenierie.c
                   |                            |om
             Status|NEEDINFO                    |RESOLVED
         Resolution|---                         |INVALID

--- Comment #7 from Philippe Mouawad <p....@ubik-ingenierie.com> ---
Closing as no feedback

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

--- Comment #3 from republic <20...@qq.com> ---
i really expect this solved; best wishes

jmeter version 3.2.x

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

--- Comment #2 from republic <20...@qq.com> ---
i thought that "CSV Data Set Config" element read file content to a list; but
all user thread get data from this list,so there must be competition.

i thought this can explain why my handcode threads load test get 14000qps, but
jmeter get only 6300qps.

my handcode threads initialized with thread's own data, no competition;

my solution was given at bug description

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

--- Comment #4 from Felix Schumacher <fe...@internetallee.de> ---
In my tests, I got over 400.000 qps using a CSV data source and a debug
sampler. It didn't matter whether I used 8 threads, 100 threads or 1000
threads.

Could you be more specific and share a test, that you use (it should be
minimalistic)?

I even looked into the FileServer implementation (that class gets used, when
you configure a CSV data source) and changed the locking to be a bit finer
grained. It didn't change the number of qps. It made the max duration a bit
lower. I will attach the patch, so you can check, if it helps in your setup.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

Philippe Mouawad <p....@ubik-ingenierie.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|JMETER_5.2                  |---

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 63639] give you three times qps improve

Posted by bu...@apache.org.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63639

Felix Schumacher <fe...@internetallee.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 OS|                            |All
             Status|NEW                         |NEEDINFO

--- Comment #1 from Felix Schumacher <fe...@internetallee.de> ---
There is not enough information here for me to understand, what your problem is
or how you suggest, we should solve it.

I suspect that you should discuss your problems on JMeter users mailing list
first.

-- 
You are receiving this mail because:
You are the assignee for the bug.