[OSM-dev] tiles at home disk usage

Frederik Ramm frederik at remote.org
Wed May 2 14:44:06 BST 2007


> Do you count the right column? I count 66GB.

Oops. I did count the right column but made another stupid mistake.

> A tile which is fully the same colour is 571 bytes. That can also be
> green tiles from parks etc, but I guess most of them is water. I count
> 12,698,593 of them. That would be 6.5G in size. But if they have to be
> stored in 512 bytes sectors, that would mean 1024 bytes of diskspace
> used per tile, making it 13G in total.

So these 13G can be expected to be removed if we do it with symlinks  
(which, to my knowledge, don't use a full sector as they sit in the  
filename space of the directory entry) or directly through the database.

Still, on a 160 GB drive, the difference between 57% and 92%, would  
be something like 55 GB, and not only those 13 GB.

> I've attached the script to generate the numbers. It's quickly hacked
> up and could use some refactoring however...

I ran the script against a statistics file I saved on 19 April,  
shortly before the introduction of blue oceans.

Total number of tiles then: 11.5 million, now: 27.5 million (up 16  
Total PNG size then: 49.5 GB, now: 66 GB (up 16.5 GB net)

If it is true that we have a sector size of 512 bytes, each tile will  
on average waste 256 bytes, and by having 16m more tiles, that wastes  
an extra 4 GB, in addition to the 16.5 GB more payload. Still, 20 GB  
more on a 160 GB harddisk is a 12.5% difference which hardly can  
account for a jump from 57% full to 92% full!

The increase seems to be part due to the maplint layer

Of the 16 million new tiles, 13.5 million have been created on the  
default map layer and 2.5 million on the maplint layer, the majority  
of the 2.5m tiles being at level 17 (still "catching up" because at  
some point I changed it to go up to level 17 instead of stopping at 16).

Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00.09' E008°23.33'

More information about the dev mailing list