On Wed, Jan 27, 2010 at 04:40:43PM -0700, Mike Swope wrote:
> In the last 2 weeks there have been hundreds of emails on the group, and
> much of it dealing with openstreetmap and imagery, and using this to help
> people on the ground.
> 
> One thing that did strike me is the need for a project such as
> openaerialmap. There was probably a TB of imagery put up for download, so
> people could use it for relief work.

Far more, actually. In total, there has been about 2.5TB of source imagery
so far, ignoring LWIR, SWIR, MWIR, and topos.

20M deltastate
1004G   digitalglobe
564M    EROS
194G    geoeye
32G google
1.3G    HDDS
31M Navy
869G    noaa
644G    worldbank-rit

Combining this source imagery with processed (warped+overlayed imagery), 
this brings us up to 6.3TB so far, and more still coming. 

http://haiticrisismap.org/?zoom=8&lat=19&lon=-72.2&layers=basephoto
http://haiticrisismap.org/?zoom=11&lat=18.58007&lon=-72.31467&layers=basephoto

> But it does seem that a nice open system would be very beneficial for
> people.
> 
> So I'm curious, what are your thoughts? How can the OAM project be used for
> a crisis?

Actually, what this has shown me is one thing, and one thing only:
we need a lot more disk than we have.

Ignoring source imagery, which can be compressed and stashed somewhere
else, and assuming that we have enough CPUs that it somehow becomes
reasonable to compress the processed data while still keeping it on 
fast disks, we've got a situation where we have the need to store 
1-2TB of imagery. For coverage of one week. For a country smaller than most
US states. With most areas not being covered with imagery of higher resolution
than a meter, and not including any additional bands.

If you *really* want to do OAM, you're gonna need better than that. Under
the situation where we're not pre-caching everything, we need to
have access to dozens or hundreds of terabytes, of disk, and the
ability to grow that amount rapidly. 

If you want to do this the Google way -- pre-cache everything -- you 
still have issues not that much different than this, if you want to solve
the 'imagery in the past' problem as well. The only difference is that you
can slightly more easily distribute your disk space, because there's no
need for CPU where the tiles are. Hwoever, I think that is likely impractical
for something like OAM using the publicaly available resources at
this time.

If the OpenAerialMap community is actually interested in going anywhere,
the first question should be "Who is going to provide lots of disk 
space that we can put near lots of CPU?" You can imagine, for the time
being, that we have the CPU, and the rackspace nearby it. The question
then becomes "Who is willing to donate lots of fast disks that can be
racked up, and how much does it add up to?"

When the potential donations start to cross the 100TB mark, we have a
potential chance of being able to get OAM started in a serious way.
Until we have that, as soon as we open the doors for imagery, we're
going to run out of space.

Best Regards,
-- 
Christopher Schmidt
Web Developer

_______________________________________________
talk mailing list
[email protected]
http://openaerialmap.org/mailman/listinfo/talk_openaerialmap.org

Reply via email to