[OSM-dev] scaling

Frederik Ramm frederik at remote.org
Tue Jan 11 08:44:41 GMT 2011


Serge,

On 01/11/11 03:28, Serge Wroclawski wrote:
> I think we're all in general agreement that the status quo won't work
> for a whole lot longer.

My assessment - and I'm not a database or sysadmin guru in the least, 
but not being a lawyer has never kept me away from legal-talk either - 
is that by simply beefing up the database server in keeping with demand, 
and adding 1-2 simple read-only mirrors next to it, we can still easily 
grow by a factor of 5 or 10.

Now how long that lasts us is of course a different question; if we grow 
exponentially by a factor of 2 per year then it'll last us three years. 
If we grow by a factor of "yearly statistics are now weekly", it'll last 
us three months.

And of course it is always good to have a working solution for "the day 
after tomorrow" already thought out. Then again nothing ages faster than 
computer stuff, and depending on when we will actually need it, what 
looks like a sensible way of scaling now might be "sooo 2010" in 2013 
(after Google have rolled out the next cool thing in distributed databases).

The key to all this is probably estimating the most probable rate of 
growth, and making plans for that.

I would also suggest to refrain from making plans for anything beyond a 
planning horizon of 24 months in OSM; it's very likely a waste of time.

Bye
Frederik



More information about the dev mailing list