EXECUTIVE SUMMARYThe workshop was organized to provide a forum for discussions focused on issues pertaining to extremely large databases. Participants represented a broad range of scientific and industrial database-intensive applications, DBMS vendors, and academia.The vast majority of discussed systems ranged from hundreds of terabytes to tens of petabytes, and yet still most of the potentially valuable data was discarded due to scalability limits and prohibitive costs. It appears that industrial data warehouses have significantly surpassed science in sheer data volume.Substantial commonalities were observed within and between the scientific and industrial communities in the use of extremely large databases. These included requirements for pattern discovery, multidimensional aggregation, unpredictable query load, and a procedural language to express complex analyses. The main differences were the availability requirements (very high in industry), data distribution complexity (greater in science due to large collaborations), project longevity (decades in science vs. quarter-to-quarter pace in industry) and use of compression (industry compresses and science doesn't). Both communities are moving towards parallel, shared-nothing architectures on large clusters of commodity hardware, with the map/reduce paradigm as the leading processing model. Overall, it was agreed that both industry and science are increasingly data-intensive and thus are pushing the limits of databases, with industry leading the scale and science leading the complexity of data analysis.Some non-technical roadblocks discussed included funding problems and disconnects between vendors and users, within the science community, and between academia and science. Computing in science is seriously under-funded: the scientific community is trying to solve problems of scale and complexity similar to industrial problems, but with much smaller teams. Database research is under-funded too. Investments by RDBMS vendors in providing scalable multi-petabyte solutions have not yet produced concrete results. Science rebuilds rather than reuses software and has not yet come up with a set of common requirements. It was agreed there is great potential for the academic, industry, science, and vendor communities to work together in the field of extremely large database technology once the funding and sociological issues are at least partly overcome.Major trends in large database systems and expectations for the future were discussed. The gap between the system sizes desired by users and those supported cost-effectively by the leading database vendors is widening. Extremely

In CIDR 2009, we presented a collection of requirements for SciDB, a DBMS that would meet the needs of scientific users. These included a nested-array data model, science-specific operations such as regrid, and support for uncertainty, lineage, and named versions. In this paper, we present an overview of SciDB's key features and outline a demonstration of the first version of SciDB on data and operations from one of our lighthouse users, the Large Synoptic Survey Telescope (LSST).