Object Oriented Database Learning Center

Objectivity Home - Government - Objectivity/DB - Webinars - Download Software

Table of contents:

Object Oriented Database vs Relational Database

CERN
(Data Gathering and Analysis)

CERN is the leading worldwide consortium, headquartered in Geneva, Switzerland, for performing scientific experiments at the sub-particle level. For their new generation accelerator, they need to build the largest database ever built, which is expected to grow to 100 PB = 100,000 TB = 100,000,000 GB = 10^17 Bytes. After evaluating all other available DBMS technology, both RDBMS and ODBMS, they concluded that the only system scalable to this level is Objectivity/DB.

Their new accelerator for High Energy Physics (HEP) is approximately 28 km in circumference. When live, data will be pumped into the database at the rate of 100 MB/sec. Such data w ill store events, which are particle collisions and their results. Various users, around the world, will perform analyses of this data. One primary goal is to seek the Higgs boson, postulated to be responsible for creating mass (otherwise the universe would be all energy). Probing deeper and deeper into nature requires more energy, larger accelerators, and, because the desired events are more rare, larger amounts of data with more sophisticated analysis.

Although the new accelerator will not be online until approximately 2003, CERN is already using the ODBMS for several past experiments, and have measured up to 80MB/sec data input, with sizes up to a TB.

Size and distribution were the key driving needs that lead to ODBMS. The natural modeling of objects for the complexity of events was also quite useful. In size, no central-server architecture could possibly handle such large sizes. These older architectures scale only as far as the largest computer available, and there are no computers that could handle such huge volumes of data and concurrent access, even if they were affordable. Instead, the CERN team required a system which could use distribution and a single logical view to transparently allow adding multiple servers, multiple disks, to support many users and many databases simultaneously. Distribution is also important for access, because the users of the database are scattered worldwide, with access via the Internet. Here, the key was the distributed ODBMS ability to selectively replicate certain subsets of all the databases and automatically re-synchronize them.

For more information, see: