<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Thu, Dec 22, 2016 at 5:20 PM, Matthias Kuhn <span dir="ltr"><<a href="mailto:matthias@opengis.ch" target="_blank">matthias@opengis.ch</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span>On 12/22/2016 05:06 PM, Alessandro Pasotti wrote:<br>
> On Thu, Dec 22, 2016 at 4:10 PM, Matthias Kuhn <<a href="mailto:matthias@opengis.ch" target="_blank">matthias@opengis.ch</a><br>
</span><span>> <mailto:<a href="mailto:matthias@opengis.ch" target="_blank">matthias@opengis.ch</a>>> wrote:<br>
><br>
> Hi Alessandro<br>
><br>
> ><br>
> > In Lyon we also talked about how to share caches between different<br>
> > server processes, do you think that the shared memory segment would be<br>
> > appropriate in that case?<br>
><br>
> What are the caches used for?<br>
><br>
><br>
> mainly per-project capabilities.<br>
><br>
> With QGIS 3 refactoring the QgsProject class will probably hold all the<br>
> information we need instead of parsing the project XML like we are doing<br>
> now, I think that we'll need to cache the QgsProject instance itself if<br>
> we want scalability by spawning multiple server processes without the<br>
> need to re-parse the project file, this in case we need to serve<br>
> multiple projects of course.<br>
<br>
</span>Is there some profiling information available about what's the slow part<br>
(parsing, loading of layers, loading of print composers...)?<br></blockquote><div><br><br></div><div>I think Patrick was working on profiling.<br></div><div>BTW, the thing are probably going to change significantly after the refactoring that will use the new QgsProject.<br><br></div><div> <br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Saving a whole project including layers etc. in a shared memory sounds<br>
like a challenge.<br>
It's possibly easier to identify which parts are slow and cache them on<br>
some backend.<br>
If this mechanism is abstract enough to work with a file based format<br>
like sqlite or memcached/redis, that would be awesome because QGIS<br>
desktop could also benefit and depolyment on a server will probably be<br>
easier if it does not depend on additional services.<br>
</blockquote></div><br><br clear="all"><br>-- <br><div class="m_3119060034557705223gmail_signature" data-smartmail="gmail_signature">Alessandro Pasotti<br>w3: <a href="http://www.itopen.it" target="_blank">www.itopen.it</a></div>
</div></div>