The TileDB Array Storage System

Co-sponsored by: Young Professionals

In this talk I will present a new system for scientific data storage called TileDB. Contrary to most traditional data management systems, TileDB models data natively as multi-dimensional arrays, which can be either dense (where every array element stores a value) or sparse (where the majority of elements are empty). I will initially present the array model and the internal mechanics of the TileDB array storage manager. I will then focus on two of the many applications of TileDB, specifically, a genomics customization, called GenomicsDB, and a geo-spatial use case. Finally, I will outline our future vision for commercializing TileDB, namely the way we envision to perform high performance analytics in a massively distributed, scalable fashion.

Speaker(s): Stavros Papadopoulos,

Location:
Room: G-449
Bldg: MIT 32
32 Vassar Street
Cambridge, Massachusetts
02139