Transcript PPT - Open Access Repository

Database requirements for the
Inner Silicon Tracker in BTeV
First thoughts
D. Menasce, S Magni - I.N.F.N. Milano
The Inner Silicon Tracker will presumably rely on the following categories of
Databases to keep track of it’s inner status:




Detector construction tracking
Configuration/Initialization constants
Calibration constants
Monitor values and reference plots
In this memo we will try to give a first look at what we envisage/expect to be
relevant for each of these categories, with particular emphasis to:





Purpose of the database
Relationship between databases
Source of the data
Expected size
Anticipated use and modes of access
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
1
Purpose of the databases
 Detector construction tracking
 Keeps track of parts of the detector:
 elements already assembled, quality test results
 stocks ordered, components tested, general inventory
 spare parts, ...
 Configuration/Initialization constants
 Holds the current status of all quantities needed to properly initialize the system and
make it ready for run or for calibration
 logical addresses of strips, suppression masks, trigger masks
 threshold values, high voltage settings and current limits...
 Calibration constants
 Will contain historical archive of calibration data (raw histograms and/or summary)
 threshold curves
 setup values used to collect current calibration data (data or pulser), ...
 Monitor values and reference plots
 Data collected at periodic intervals to check the status of the detector
 profile histograms, occupancy plots, residuals
0
 crude track segments reconstruction statistics, physics signals ( K s ?)
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
2
Relationship between databases
 Detector construction tracking
 May have a relationship with all databases to define which components are active.
 Configuration/Initialization constants
 Gets updated once a new calibration indicates there is a need to change threshold values.
 Definite relationship with the Calibration Constants database, source of data needed
to compute new set of initialization constants
 Possible relationship with the Monitor Values database: the latter gets here reference
values to check against monitored ones
 Possible relationship with the Detector Construction database: initialization might be
done only on components declared installed by the latter database.
 Calibration constants
 Definite relationship with the Configuration Constants database. Values of calibration
are used to compute threshold, masks to kill noisy/dead channels used by the latter db.
 Possible relationship with the Detector Construction database: calibration might be
done only on components declared installed by the latter database.
 Monitor values and reference plots
 Possible relationship with the Configuration Constants database. Values of calibration
might be used as reference values for quantities needed to operate monitor.
 Possible relationship with the Detector Construction database: monitor might be
done only on components declared installed by the latter database.
9 December 2003
D. Menasce. S. Magni: Database
3
requirements for the Silicon Tracker
Source of the data
 Detector construction tracking
 Main source of the data will presumably be a WEB interface, to allow collaborators in
charge of assembly of the detector to enter values remotely.
 Should large numbers of data be available as files (e.g. preamplifier chips shipped by
a foundry, scripts or programs will transfer them into the database)
 Configuration/Initialization constants
 Most part of the data (thresholds, kill/trigger masks) will be entered by specialized
programs. These will get data from Calibration database and compute relevant quantities.
 Small part will be entered from WEB interface.
 Calibration constants
 Source will be from programs gathering data from DAQ.
 Monitor values and reference plots
 Source will be from programs gathering data from DAQ.
These databases will be populated at low I/O rate, no direct feed from DAQ in real time is
envisaged (no time-critical I/O).
At any rate, databases are known to be slow in write mode, but quite fast in read; the most
frequent anticipated mode of access will be read (particularly true for WEB access)
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
4
Expected size
 Detector construction tracking
 Precise size of this database is difficult to predict.
 Rough estimate is a few hundred Mb.
 Configuration/Initialization constants
 The number of strips in the detector is of the order of 3x105 (36 X 36 cm2 option)
 For each strip logical address etc.. are recorded (static information, no historical archive)
 There are ~2300 chips (128 channels each) each holding a threshold value; this
information is time dependent and will be updated presumably (pessimistic) once a
month. Biases, current limits etc… (for each chip or for each detector wafer) will be
updated much less frequently.
 Rough estimate is a few Mb per month.
 Calibration constants
 Threshold and noise values are the main content of this database. It has not yet been
decided whether these data will be stored as raw (threshold curves, histograms) or as
simple threshold/noise pairs.
 Opting for the most conservative choice (complete information as histograms)
3x105 strips x 100 bin/hist x 4 bytes/word = 120 Mb.
 Considering a weekly update frequency estimate is ~500 Mb/month.
 Monitor values and reference plots
 Difficult to predict, depending on type of quantity stored (simple numerical values or
plots/distributions). Assuming profile/occupancy histograms, 3x105 strips x 4 bytes/word
we get ~1.2 Mb. Assuming 10 possible quantities monitored per strip, we get an
assumed size of about 20 Mb per shift (8 hours?).
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
5
Anticipated use and access patterns
 Detector construction tracking
 Use will be through WEB forms and reports to allow people to track progress in
commissioning, installation and testing of the complete detector.
Programs will read these data as well to gather information of existing/available
components (calibration and monitor are meaningful only for installed parts).
 Of no lesser usefulness will be the disposal of an historical archive of each
sub-component origin and handling. Spare parts location is also of consideration.
 Areas of use will therefore be: WEB, online systems
 Configuration/Initialization constants
 This database contains all the information needed to properly initialize the detector.
 Data will be accessed both by detector initialization program and by WEB scripts
to allow people on shift to check out what the current image of detector looks like.
 Areas of use will therefore be: WEB, online systems
 Calibration constants
 Areas of use will therefore be: online systems
 Monitor values and reference plots
 Monitor data will be used both to decide upon need of a calibration or an update
of initialization constant values, and as source of timelines used by offline programs.
 Montecarlo simulation needs to be knowledgeable about run periods shifts in data
characteristics, information gathered in the form of timelines.
 Area of use will therefore be: WEB, online systems, offline programs/data analysis.
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
6
Considerations about the DBMS (1)
 Which DBMS meets the requirements specified for the Silicon tracker?
 Given the anticipated sized and foreseen uses, most likely a open/source DBMS is
appropriate. MySQL, miniSQL and POSTGRES are all good candidates: reliable, easy
to use, full fledged relational databases.
 miniSQL no longer developed
 MySQL currently the most used
 POSTGRES developed with inferential capabilities: update of tables can be triggered
by pre-specified conditions. Algorithms moved from interface program to the database
itself, with centralization of procedures.
 Since anticipated use is, without exception, through specialized programs, which
API is offered by the DBMS is of consideration. Those indicated above all have :
 C and Perl APIs
 What about backups?
 Backups are of course essential. A general solution should be devised to encompass all
databases in use or foreseen by the collaboration.
 Possible solution could be mirroring of the database at off-site institutions
(overnight update of the mirror sites such as is done with offline libraries)
 Any connection with other databases?
 Most likely: mechanical mounting of the Silicon tracker has parts in common with the
Straws detector. In the design of each database this consideration should be taken into
account.
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
7
Considerations about the DBMS (2)
 Standards are important
 It would be safe to assume that several databases will constitute the repository of
many BTeV data. Adoption of a specific DBMS and design of internals should be made
keeping in mind that standards solve cross database communication issues.
 as an example, many programmers take for granted that the DBSM keeps an internal
ID for a table (useful to join tables together). This is NOT an SQL standard: porting
data to another DBMS becomes suddenly a nightmare.
This feature, available (and actually useful) in POSTGRES is not in MySQL, which
adheres to standards more strictly
 Data representation should be independent of particular DBMS choice: this insures
maintainability of the data over long periods of time (should support for a particular
DBMS dwindle in time)
 Binary data?
 It can be sometimes useful to store binary chunks of data (GIF file drawings, histogram
files ecc…). Different DMBS’s provide specific features to this extent.
 Operating systems
 Some database could be available on Windows and some on Linux.
 ODBC and JDBC mechanisms to allow inter-communication between platforms should
be adopted. Particularly important for WEB served databases.
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
8
Geometry description database
 Purpose of this database
 This database should contain a geometrical description of the detector:
 Positions in space (derived either from technical drawings, CAD files, or from technical
surveys or even from offline alignment)
 Possibly shapes description and hierarchical relationship between components
 In the present context, this database has a special status
 A Geometry Manager is deemed necessary to de-couple the various possible simulation
codes from the actual data that represent shapes, volumes and their position in space.
XML has been envisaged as a possible unifying description language.
 The issues is whether we need a database description of the geometry from which an
XML description can be build upon request, or if XML actually IS the database.
A conventional database source for an XML description file is probably preferable since
XML descriptions can be rebuilt from there selecting upon run periods or any other
criteria.
 Architectural decisions upon this database should be postponed to
a general discussion about Geometry Manager
9 December 2003
D. Menasce. S. Magni: Database
requirements for the Silicon Tracker
9