OSGeo Tiling Spec

If you haven’t already seen the backchatter from the FOSS4G conf, there is a big push at the moment on implementing a common spatial tiling scheme.

I’m not going to bother rehashing, so swing by Distributed Tile Caching or the live docs at Tile Map Service Specification to get up to date quickly.
The distributed caching article is certainly an interesting read. Something that caught my eye today was a similar discussion on #worldwind about the futility of trying to use torrents or other technology such as coral to help distribute these caches of tiling goodness.

My question, what about Amazon S3? I have been using this service since its release for offsite backups of 60gb+ and can not praise the system enough for performance and reliability. WebDav, Torrent, SOAP … the list goes on.

Sure its not free, but it sure as hell is cheap … Based on Schuylers calculation from the wiki article,

Taking 15m pan-sharpened Landsat-7 composites as an example, at a tile size of 512 x 512 pixels, each tile would be about 7,680 meters on a side, or about .0625 degrees across. Plugging in the other values, we get a maximum of 22,118,400 tiles in the layer.

Assuming the optimum size of 64kb is reached per tile, we’re looking at 1415.577gb of physical storage. Lets take a wild guess of 50gb of transfer per month, with the actual tiles only be updated annually and we have the following,

$0.20 * 1415 = $283 to initially upload the cache

$0.15 * 1415 * 12 = $2547 for a years worth of storage

$0.20 * 50 * 12 = $120 of user transfers (eg. downloads)

Total: $2950

Cheap, no? Try it out, you wont be dissapointed … i know of at least 1 other mapping outfit that have enquired into s3 for similar tile storage for their application. I’d be keen to see a proof of concept to see whether this is do-able or not *looks at refractions*.

Certainly if my figures are correct, the costs are almost negligble even for you poor opensource developers :)

3 thoughts on “OSGeo Tiling Spec”

  1. Yeah you’re right.

    There certainly isnt anything commercially available that can compete with those prices.

    The only thing i failed to factor into the calculations is obviously the 2 way cost of someone actually uploading 1.4tb of data. In the scheme of things though, the total is still going to be very small.

Comments are closed.