It seems likely you'd have to process as you go, and not store it all. Or have one hell of a datacenter/buffer like the LHC does, holding it until it can be crunched down to a "simple" network instead of voxel data.