Migrating CRISTAL from Objectivity to RDBMS Andrew Branson
Migrating CRISTAL from Objectivity to RDBMS Andrew Branson University of the West of England CERN, 13/01/2014
CRISTAL • Production tracking system for CMS – Project started 1996 – Tracked construction of CMS ECAL Barrel, Endcap, ECAL Electronics, Preshower. – Ran at CERN and other sites, or imported data from other remote construction software • Extremely object-oriented concept. • Objectivity was an obvious choice.
CRISTAL 1 • Objectivity-based prototype (CRISTAL 1) completed 2000 • Collected prototype data: – Crystals and test ECAL modules at CERN and INFN, Rome – APD capsules at IN 2 P 3, Lyon. • C++ backend, Java UIs. • Heavily integrated with OO DB, required extensive re-engineering to migrate to alternative database.
Redesign • Prototype had design problems. • Object DBs had failed to catch on, support waning at CERN. • Large amount of work to remove Objectivity. • Small amount of production data to migrate. Ø Decision made to redesign and redevelop from scratch
CRISTAL 2 • Production system deployed August 2003 • Storage still object based, but with abstraction layer. • Plugins for Oracle, My. SQL, SQLite. • Fall-back XML file serialization.
CRISTAL 2 • Application design implemented using descriptions - metamodels. • XML (text) based. Portable, human readable. • Self-documenting, web standards (XSD) – Some preservation against technology obsolescence • Investigating migrating to XMLDB for enhanced query capability.
Conclusion Technology dependency comes in two flavours: • Dependency on Implementation – Prevent redesign through abstraction. – Small, self-contained redevelopment when obsolete. • Dependency on Philosophy – Could require large re-engineering. – Only protection is to safeguard data, schema and logic in a documented form that could be reinterpreted in the future.
End
- Slides: 8