
Hi all, I'm planning to index my spherical data in MonetDB byHEALPix. I will need it for: 1. Data partitioning - distribute the data uniformly on the DB nodes, but keep them sorted (keep points close on the sphere on the same node). 2. Neighbor lookup - can I actually index my spherical coordinates stored in MonetDB by some kind of a BTree of HEALPix pixels? I understand how computing a hash for the coordinates and indexing on it helps in relational DB, but how do the indices function in an array DB like MonetDB? I'd like to benefit from the array storage of the data, because I will often look for nearest neighbors and compute distances between these neighbors. I will also divide the data in the beginning for parallel tasks by some calculation of their pixels, but that will be done only once in the beginning, the main computation will be around clustering these objects, which means lots and lots of neighbor searches and distance computation. Has somebody already tried to use HEALPix, or some other spherical index in MonetDB? How can I control the distribution of data across the nodes (e.g. sort them by this index)? How can I use and benefit from this kind of indexing in MonetDB? Thank you very much for even some questions answered. Best regards, Jiri Nadvornik Astronomical Institute AV CR Stellar department Czech Republic nadvornik.ji@gmail.com
participants (1)
-
Jiří Nádvorník