Andy,<br><br>thank you for your valuable input.<br>Do you think, that running (not importing) the database will be still slow at EC2?<br>I liked the idea of using some kind of load balancing (one machine online, one offline for upgrading indexes) which will be easily set up with the cloning feature.<br>
Your EC2 experience doesn't sound very promising for achieving this goal.<br>Do you know another faster host which has this cloning feature?<br><br>Regards<br>Frans<br><br><div class="gmail_quote">2010/1/4 Andy Allan <span dir="ltr"><<a href="mailto:gravitystorm@gmail.com">gravitystorm@gmail.com</a>></span><br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div class="im">On Mon, Jan 4, 2010 at 2:52 PM, Brian Quinion<br>
<<a href="mailto:openstreetmap@brian.quinion.co.uk">openstreetmap@brian.quinion.co.uk</a>> wrote:<br>
<br>
>> I started the process on a standard Amazon cloud 32-bit 1-Processor Fedora unit.<br>
><br>
> Not any immediate problem but I'd guess due to the size of node IDs a<br>
> 32 bit machine has a limited time left - I'd guess around 2 years.<br>
<br>
</div>So Amazon EC2 machines have stunningly dreadful disk performance. It's<br>
a continual headache that I'm running the cyclemap on one (an<br>
extra-large instance). I use "normal" mode for the cyclemap import,<br>
but Matt used to use slim mode for his minutely tileserver and it<br>
would take about a week to just run osm2pgsql (I'll let him correct me<br>
if that's wrong). Although I haven't used the gazetteer import mode on<br>
one, I can't imagine for a minute that it'll be in any way quick.<br>
<br>
My advice would be to find another host other than EC2 if you are<br>
doing anything involving osm2pgsql (or anything else that isn't<br>
running in RAM).<br>
<br>
Cheers,<br>
<font color="#888888">Andy<br>
</font></blockquote></div><br>