<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">Hey Johannes,<br>
<br>
the preparation of the graph is quite memory and time intensive,
yes.<br>
And for enabled turn restrictions also GraphHopper will probably
need more RAM. <br>
<br>
Regarding running only the OSRM service itself: I think if you
only had 5GB and the program needs 4GB, I think 1GB could be too
few for the system itself (and your local services). But the crash
could be a bug which you should report.<br>
<br>
For GraphHopper you could try even with less RAM if you load the
graph via memory mapped data access (osmreader.dataaccess=mmap),
but this will slow down the query response time. Sometimes so bad
that it is not worth, only for memory restricted systems like
Android/RasperryPi or probably with SSD discs.<br>
<br>
Another trick could be to use the chPrepare.createAStar() method
to boost the query time a bit, but only if you accept suboptimal
paths. (precision vs. speed depends on the approximationFactor in
AStar)<br>
<br>
Regards,<br>
Peter.<br>
<br>
<br>
</div>
<blockquote
cite="mid:CAMiU5bvGAZD1Xc1RvPUiKO1uXnkWO9Q_C3-RzEzSf5QbN0Z_Zw@mail.gmail.com"
type="cite">
<div dir="ltr">
<div>
<div>
<div>Hello Peter,<br>
<br>
</div>
<div>my server is an old HP one with 5B RAM DualCore. I
tried to get OSRM running with the germany osm file from <a
moz-do-not-send="true" href="http://geofabrik.de">geofabrik.de</a>.
But finally it failed. I created the routing index files
on an amazon instance, which needed max between 10 and 11
GB RAM. The osrm-routed program just needs 3,9 GB RAM, so
i tried to run it on my old hardware. Normally it should
be enough, but the start process was killed with an alloc
error. The swap partition was not used or ignored. Very
strange.<br>
<br>
</div>
<div>So i tried the graphhopper. Indexing the same extract
file needs something between 1 and 2 GB Ram, during
runtime just 1GB, so it fullfilled my requirements. One
recordset could be calculated in 0,033 seconds, while <span
id="result_box" class="" lang="en"><span class="">parallelisation
reduces the runtime about 1/n.<br>
<br>
</span></span></div>
<div><span id="result_box" class="" lang="en"><span class="">Greeting
Johannes<br>
</span></span></div>
</div>
</div>
</div>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
<pre wrap="">_______________________________________________
GraphHopper mailing list
<a class="moz-txt-link-abbreviated" href="mailto:GraphHopper@openstreetmap.org">GraphHopper@openstreetmap.org</a>
<a class="moz-txt-link-freetext" href="http://lists.openstreetmap.org/listinfo/graphhopper">http://lists.openstreetmap.org/listinfo/graphhopper</a>
</pre>
</blockquote>
<br>
<br>
<pre class="moz-signature" cols="72">--
GraphHopper.com Your way is our destination!</pre>
</body>
</html>