Download presentation
Presentation is loading. Please wait.
Published byFelicia Jordan Modified over 6 years ago
1
The ZEUS Event Store An object-oriented tag database for physics analysis Adrian Fox-Murphy, DESY CHEP2000, Padova
2
The ZEUS Experiment ZEUS experiment at HERA, Hamburg.
Electron-proton collisions at 318GeV. Operational since 1992 In 1996, started to augment existing offline environment with a tag database (ZES) based on Objectivity/DB. First experience with OO software for ZEUS. Tag database operational since 1997. CHEP2000, Padova
3
ZEUS Offline Environment
Raw data rate: 10 Terabyte / year, 1MB/s Event rate: ~10Hz, 100Kb/event Centralised batch processing for physicists’ jobs 3 SGI Challenge XL machines Migrating soon to farm of 30 Linux PCs (M. Kowal’s talk) Raw and reconstructed data stored in ADAMO database User analysis jobs in FORTRAN. CHEP2000, Padova
4
CHEP2000, Padova
5
ZEUS Event Store (ZES) Must access required events efficiently
Formerly Calculated 128 ‘DST bits’ Made boolean selection of bits Now... Populate a database with (250) physical variables calculated at reconstruction for each event. Kinematics, electrons, muons, hadrons, jets Calorimetry, triggers Select data on the values of the physical quantities. CHEP2000, Padova
6
CHEP2000, Padova
7
ZES (contd.) Cons: Use commercial product Objectivity/DB Pros:
(v3.8 and v4.02). Pros: exists; meets our current and future requirements; tested in HEP community; OO/C++. Cons: need special interface to existing event database (ADAMO); different language depend on Objectivity (support, survival) CHEP2000, Padova
8
Database schema Objectivity persistent object hierarchy
Federated Database Database (correspond to MB files) Container (HERA Run) Basic object (Event, ADAMO File) Currently 315 databases. Data in ooVarray variable length arrays gives flexibility for expansion CHEP2000, Padova
9
Lots of glue... 18k lines of C++ provide Interfaces and tools
Interface to main ADAMO database Interface to EAZE analysis job environment Predicate string querying User command line query tool Ntuple generation tool Administration Database population consistency checking, troubleshooting
10
Selecting Events Card file interface
select run range select triggers cut on physical variables using Objectivity/DB predicate strings : ((Ee>5)and(Zvtx>-50)and(Zvtx<50) and(Eminpz>35)and(Eminpz<65)and(Yjb>0.04)) Intuitive for physicist. Precisely tailored event selection (efficient). CHEP2000, Padova
11
Populating the database
Two things required : The physical variables to load. Calculated during offline reconstruction. A pointer to the location of the event in the ADAMO database. Actually use a two-step process first store physical variables temporarily in ntuples (used also for data quality monitoring) then load ZES OO database from ntuples CHEP2000, Padova
12
Status of the ZES database
CHEP2000, Padova
13
Experience Positive response from users. Adopted as standard.
Can greatly reduce the time required to analyse the data. Factor 3-10 improvement typical. On request, the number of physical variables stored for each event was increased from 93 to 250. Problems database population must be done with care. Many things can go wrong. Consistency checking vital. CHEP2000, Padova
14
Summary ZES is an efficient and flexible tag database
Successfully integrated into ZEUS (ADAMO/EAZE). Now holds 100 Million events from 5 years. Variables expanded to 250. Jobs run faster. ZES and Objectivity are flexible and powerful enough to meet current and future requirements. OO/C++ advantages Development, maintenance HEP OO analysis and visualisation tools Future evolution into fully OO software.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.