We usually try to load the systems with as many concurrent users as we
can. The workload measures the throughput while keeping a check on
response times. The loaded data has to at least match the scale you want
to run. It can be bigger, though.
Runs at small scales are affected by the small number of samples. They
may need to run with a larger steady state in order to pass (so there
will be enough operations to have a meaningful sample size. Similarly,
small loaded data may incur the irregularities of random number
generators and applied to a limited number of instance.
Generally, runs at a scale of 200 concurrent users or more over a steady
state of 10 minutes (600 secs) or more are considered having the
critical mass for random number generators to give you predictable
results. The larger the scale, the smaller the required steady state.
But I won't keep the steady state less than 5 minutes to give you
meaningful data.
-Akara
Shahnavaz Valayil wrote:
Thanks for your help. I tried with load scale 200 and it passed. What
are the criteria for choosing the correct load scale count(and
concurrent users), and the time(ramp up, steady state,ramp down, Time
between starts). All this time i was testing this on a VM and now i am
planning to run this on server with more then 32GB RAM running CentOS.
Thanks again.
Shahnavaz
On Tue, Nov 25, 2008 at 11:30 AM, Akara Sucharitakul
<[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:
The load scale (under DB server) is currently the bare minimum of
25. I'd load for at least 100 to 200 concurrent users, even if you
only drive with 25 concurrent users. You're already very close to
passing, though.
-Akara
Shahnavaz Valayil wrote:
Thanks for moving the discussion to olio list.
I have tried 25 and 50 scale and non of them worked for me. Does
the ramp up/steady state/ramp down time also matters? I have
attached the run.xml file from my last test. Please let me know
if you need more info.
Thanks
Shahnavaz
On Tue, Nov 25, 2008 at 10:57 AM, Akara Sucharitakul
<[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
<mailto:[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>> wrote:
Can you please let me know the scale (number of concurrent users)
you're loading the data and running this? Either of these can be
contributing factors for such failed results. This is due to too
small of a data set.
I'd first try to load at a larger scale in any case.
Yes, this is an Olio discussion. Project Olio is here:
http://incubator.apache.org/olio/. The alias to use is
[email protected]
<mailto:[email protected]>
<mailto:[email protected]
<mailto:[email protected]>>. Thanks.
-Akara
[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
<mailto:[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:
Whenever i run the performance benchmark i always get the
Failed
result. I dont see any error in any of the log file. The only
fail item in the summary is under Miscellaneous
Statistics. It
says "Average images on Tag Search Results 6.15 >= 7
FAILED".
Why is it failing? Do you need to tweak some parameter to
make
it pass? I tried many values under Driver tab for
RampUp,Steady
State and Ramp Down but the result is always "fail".
I am using CentOS 5.2, Faban 091008 and latest olio.
Thanks for
your help.
Shahnavaz
P.S: Sorry, I dont know how to move this under olio(if it is
olio issue), i dont see it anywhere in sunsource.net
<http://sunsource.net>
<http://sunsource.net> site.
------------------------------------------------------
http://faban.sunsource.net/ds/viewMessage.do?dsForumId=111&dsMessageId=89809
<http://faban.sunsource.net/ds/viewMessage.do?dsForumId=111&dsMessageId=89809>
<http://faban.sunsource.net/ds/viewMessage.do?dsForumId=111&dsMessageId=89809
<http://faban.sunsource.net/ds/viewMessage.do?dsForumId=111&dsMessageId=89809>>
To unsubscribe from this discussion, e-mail:
[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
<mailto:[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>].
