Tryton 4.2 is slow compared to previous versions

[appu rongala]

Hi Devs,

I have installed tryton4.2  and i feel it is slower than previous versions.

I have tryton3.4 installed and that is faster than 4.2.

Am i doing anything wrong?

Thanks in Advance

-Appu

[Sergi Almacellas Abellana]

El 25/05/17 a les 21:42, appu rongala ha escrit:

> Hi Devs,
>
> I have installed tryton4.2 and i feel it is slower than previous versions.
> I have tryton3.4 installed and that is faster than 4.2.
> Am i doing anything wrong?

I don't know, but what is "slower" or "faster". Could explain which
actions are you doing? Do you know how much slower?

Could you explain a little bit, how is your setup? i.e: Databse, Server
OS...



--
Sergi Almacellas Abellana
www.koolpi.com
Twitter: @pokoli_srk

[Raimon Esteve]

Hello,

2017-05-25 21:42 GMT+02:00 appu rongala <appu.r...@gmail.com>:
> Hi Devs,
>
> I have installed tryton4.2 and i feel it is slower than previous versions.
> I have tryton3.4 installed and that is faster than 4.2.
> Am i doing anything wrong?

Yes, since 4.0 is more slower than 3.8, 3.6... when have big data or
complex forms. We tested with bin/trytond and WSGI app.

At the moment we think it is the werkzeug [1] and requests.

Finally we tested with a WSGI comiled with C and timeit reduce 50-60%.
At the moment, we are still investigating....

Regards.

[1] http://werkzeug.pocoo.org/

[Cédric Krier]

On 2017-05-26 14:09, Raimon Esteve wrote:
> Hello,
>
> 2017-05-25 21:42 GMT+02:00 appu rongala <appu.r...@gmail.com>:
> > Hi Devs,
> >
> > I have installed tryton4.2 and i feel it is slower than previous versions.
> > I have tryton3.4 installed and that is faster than 4.2.
> > Am i doing anything wrong?
>
> Yes, since 4.0 is more slower than 3.8, 3.6... when have big data or
> complex forms. We tested with bin/trytond and WSGI app.

Without any standard measurements, we can not say anything.
If you have such data, please share it. And also the tools to make the
measures.
We also need to know what is measured? Tryton is a complex software with
many usages. It does not make sense to just say "it is slow".

--
Cédric Krier - B2CK SPRL
Email/Jabber: cedric...@b2ck.com
Tel: +32 472 54 46 59
Website: http://www.b2ck.com/

[Raimon Esteve]

Hie Cédric

2017-05-26 14:57 GMT+02:00 Cédric Krier <cedric...@b2ck.com>:
> On 2017-05-26 14:09, Raimon Esteve wrote:
>> Hello,
>>
>> 2017-05-25 21:42 GMT+02:00 appu rongala <appu.r...@gmail.com>:
>> > Hi Devs,
>> >
>> > I have installed tryton4.2 and i feel it is slower than previous versions.
>> > I have tryton3.4 installed and that is faster than 4.2.
>> > Am i doing anything wrong?
>>
>> Yes, since 4.0 is more slower than 3.8, 3.6... when have big data or
>> complex forms. We tested with bin/trytond and WSGI app.
>
> Without any standard measurements, we can not say anything.
> If you have such data, please share it. And also the tools to make the
> measures.
> We also need to know what is measured? Tryton is a complex software with
> many usages. It does not make sense to just say "it is slow".

start a client action -> trytond -> finally the response client.

I'm agree with Appu about when compare 4.x is slow that a minor version.

Regards

[Cédric Krier]

This is not a scientific measurement.

[Sergi Almacellas Abellana]

El 26/05/17 a les 15:10, Raimon Esteve ha escrit:

On which client actions? How you are running both servers? Which
database server? How long takes to process a request on each version?
How many requests have you executed? Is the first request or the other
ones?

There are too much factors that can influence, so if you do not provide
meaningful examples, it's useless.

[Maximilian Holtzberg]

You can try to run the client over mitmproxy like this:
mitmproxy -R http://localhost:8000 -p 8088
and connect the client to it. It gives you some timing informations
like:

First request byte 2017-05-26 12:07:21.961
Request complete 2017-05-26 12:07:21.985
First response byte 2017-05-26 12:07:22.027
Response complete 2017-05-26 12:07:22.043

[Sergi Almacellas Abellana]

El 26/05/17 a les 17:51, Maximilian Holtzberg ha escrit:

You are still not proving any evidence that trytond is slower.
And also don't providing details about the setup-up.

[Nicolas Évrard]

* Sergi Almacellas Abellana [2017-05-29 10:07 +0200]:

>El 26/05/17 a les 17:51, Maximilian Holtzberg ha escrit:
>>

>>You can try to run the client over mitmproxy like this:
>>mitmproxy -R http://localhost:8000 -p 8088
>>and connect the client to it. It gives you some timing informations
>>like:
>>
>>First request byte 2017-05-26 12:07:21.961
>>Request complete 2017-05-26 12:07:21.985
>>First response byte 2017-05-26 12:07:22.027
>>Response complete 2017-05-26 12:07:22.043
>>
>You are still not proving any evidence that trytond is slower.
>And also don't providing details about the setup-up.

I think Max is just explaining to Raimon how he can get some
measurements in a more reliable way regarding interaction with the
server then "I click on the button and it feels longer then before".

--
Nicolas Évrard - B2CK SPRL
E-mail/Jabber: nicolas...@b2ck.com

[Jan Grasnick | Grasbauer UG]

Am 29.05.2017 um 11:58 schrieb Nicolas Évrard:
> * Sergi Almacellas Abellana [2017-05-29 10:07 +0200]:
>> El 26/05/17 a les 17:51, Maximilian Holtzberg ha escrit:
>>>
>>> You can try to run the client over mitmproxy like this:
>>> mitmproxy -R http://localhost:8000 -p 8088
>>> and connect the client to it. It gives you some timing informations
>>> like:
>>>
>>> First request byte 2017-05-26 12:07:21.961
>>> Request complete 2017-05-26 12:07:21.985
>>> First response byte 2017-05-26 12:07:22.027
>>> Response complete 2017-05-26 12:07:22.043
>>>
>> You are still not proving any evidence that trytond is slower.
>> And also don't providing details about the setup-up.

I had the same issue with 4.2 and replaced werkzeug with cherrypi. Since
it happened in a very strange environment (windows, trytond in cygwin) I
didn't give it much thought. But the improvement was significant (from
mostly unusable to wow :) . I don't know your environment but you can
give it a try and report back your results . So we have an entry point
for further investigations.

[Sergi Almacellas Abellana]

El 29/05/17 a les 13:58, Jan Grasnick | Grasbauer UG ha escrit:

trytond default comandline uses werkzeug run_simple [1] which as
explained on their docs [2] it performs poorly under high load. Indeed,
there is some thread on this list [3] explaining how to configure it as
WSGI server, which will provide better performance.

Hope it helps.


[1] http://hg.tryton.org/trytond/file/93b914901108/bin/trytond#l46
[2] http://werkzeug.pocoo.org/docs/0.12/serving/#werkzeug.serving.run_simple
[3] https://groups.google.com/d/msg/tryton-dev/GI2h2cSUG6A/e2F2H3nXBQAJ