The problem with serving large datasets from multiple machines is you would then need to have each of the filesystems mounted around to the other machines and a nasty web of symlinks to keep it all together and served which IMO is unworkable... or ... use something like hdfs ... which I don't think is well suited to serving data from. These small machines would work well for small datasets, but I believe we need to start with some large base datasets that are larger than 4TB in size ... thus the need for at least large machine to start with.
Jeff On Wed, Dec 9, 2009 at 5:22 PM, Anthony <[email protected]> wrote: > On Wed, Dec 9, 2009 at 12:41 PM, Jeffrey Johnson <[email protected]> wrote: >> >> 1) Buy/Build a machine with a decent amount of disk to get this thing >> off the ground. Once built, it would require somewhere to host it, but >> that could be less of a problem than putting the machine together at >> first. By 'decent' ... I mean at least 10TB probably as much as 20 or >> 30 just to get the ball rolling here, otherwise we quickly >> (immediately) run into resource issues. This kind of machine (30TB >> disk) can be built for roughly $7.5-10k. It seems to me that this kind >> of $ could be raised if we had an appropriate plan. > > What if we got 8 machines with 4 terabytes each, or 4 machines with 8 > terabytes each? What kind of CPU would we need in each? > _______________________________________________ talk mailing list [email protected] http://openaerialmap.org/mailman/listinfo/talk_openaerialmap.org
