PHENIX VTX DB Overview Peter Steinberg Rachid Nouicer
PHENIX VTX DB Overview Peter Steinberg Rachid Nouicer, Burt Holzman Brookhaven National Laboratory December 9, 2004
PHOBOS Experience • PHOBOS chose a single DB scheme early on • Used it comprehensively for a variety of information • Si testing • C-V curves etc. • Module assembly • Calibrations, quality, etc. • Detector operation • Pedestals, Gains, etc. • Same DB also used for other subsystems, slow controls, RHIC data, etc.
“Golden Rules” (Bolek Wyslouch) Ø One central DB Ø Avoid distributing among institutions Ø DB must be robust Ø Commercial products have professional support, are optimized (so faster), etc. Ø Log all actions, even failures Ø Must be part of internal Table design Ø Data visualization must be available ASAP Ø All first attempts at data entry usually fail – check everything by eye! Ø Do not reuse another DB implementation Ø Leverage experience, not actual DB code Ø DB and web interface should be separated Ø Each should be replaceable independent of other
PHOBOS Si Testing Roadmap UIC: Rachid Nouicer Burt Holzman Mike Reuter MIT/BNL: Bolek Wyslouch UIC Si Testing ORACLE Server ANL (module testing): Birger Back Alan Wuosmaa MIT: Gerrit van Nieuwenhuizen Kris Gulbrandsen Carla Vale Patrick Decowski MIT Si Testing Online: PHOBOS CH Offline: RCF, all institutions, etc.
Server History Ø Silicon Testing and Assembly v 1997 Purchased Windows NT server Pentium Pro 200 MHz, installed Oracle 7. 3 (~2 Gbyte) Ø Main Experimental Database (~ 30 Gbyte) v 1997 Windows NT server, 500 MHz, 18 Gbyte disk (decommissioned) (Oracle 8. 0. 5) v 2001 Windows NT server 600 MHz, 36 Gbyte + 300 Gbyte RAID (Oracle 8. 0. 5) v 2003 DELL Power. Edge 2650, Dual 2. 8 GHz, Linux, RAID (Oracle 9. 2. 0. 4. 0)
PHOBOS Implementation • Oracle chosen in 1996 and used today • Upgraded occasionally • DB Machine upgraded consistently • 200 MHz NT Dual 2. 8 GHz Linux • DB size moderate • 10’s of GB – similar to PHENIX DB • No mirroring • Was tried early on and abandoned • Problems solved by faster networks
Immediate Feedback • Test data goes in from Labview • Should come out immediately on web
PHOBOS “Portal” Keep everything on one main page (avoid nested links) Done with MS Front. Page (ASP) & Oracle
Scope for PHENIX • VTX project will have several stages • Wafer testing • BNL, SUNYSB, UNM, RIKEN • Assembly & Integration • BNL • Operations • PHENIX • All information gathered at each stage must be persistent and accessible to next stage
Configuration for Wafer Testing RIKEN/ BNL Sensor inventory, history properties SUNYSB UNM PHENIX VTX DB Module history, inventory, properties Ladder history, inventory, properties RIKEN/ JAPAN Detector history, inventory, properties
PHENIX VTX DB • PHENIX Experience points to using POSTGRES • No particular reason to use Front. Page/ASP, etc. • Can move to Apache/PHP • Common decision these days • The only tricky issues involve • Configuration • Person-power
Data Types • All modern SQL DB’s support many types of data • At testing phase, emphasis placed on raw sensor values (CV IV curves etc. ) • During operations, emphasis should be on the system • Easier to work with objects, arranged geometrically • PHOBOS uses ROOT objects in BLOBS
Proposal for DB Testbed • Main PVD machine • • • Similar to current PHOBOS setup DELL Poweredge rack system Dual CPU, O(100 GB local disk) ~1 TB RAID Array (400 -1000 GB disk) POSTGRES DB Apache webserver running PHP (ASP? ) • Secondary machine • Backup & mirror • Study failover techniques • Each machine ~$5000, so requesting $12 k • Of course can start with one machine • Can also scavenge RCF castoffs • BUT: this machine may well be the production machine – early investment may pay off
Manpower • We have identified several sources • BNL Chemistry • Rachid – expert silicon user • Burt & Peter – past & present PHOBOS computing czars, silicon demi-experts • Bolek will consult on design choices • “Mainstream” PHENIX • Irina – PHENIX DB czar • Dave – PHENIX computing czar • Abhay – can offer student labor
Reasonable Short-Term plan • Provide repository for wafer testing thorughout VTX collaboration • Problem reduces to a few tables • While fully-loaded testbed machine would be ideal, not in principle necessary • Provisional DB could be migrated from a less powerful machine (e. g. RCF salvage) • People: Irina, Rachid, Peter, Burt
Medium-Term plan • Start working towards DB which will be used for VTX integration • Full set of testing data • Hybrid information • Module information • Detector information • Time-dependent information (e. g. slow monitoring) • Will require broader input from various parties • Existing PHOBOS setup can serve as useful template
- Slides: 16