history plotting tool for data quality monitoring
TRANSCRIPT
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 617 (2010) 263–265
Contents lists available at ScienceDirect
Nuclear Instruments and Methods inPhysics Research A
0168-90
doi:10.1
� Corr
E-m1 O
journal homepage: www.elsevier.com/locate/nima
History plotting tool for Data Quality Monitoring
D. Giordano a,1, A.-C. Le Bihan c,1, A. Pierro b,1, M. De Mattia d,�,1
a CERN, Switzerlandb INFN-BARI, Italyc Institut Pluridisciplinaire Hubert Curien Strasbourg, Franced INFN and University of Padova, Italy
a r t i c l e i n f o
Available online 19 July 2009
Keywords:
CMS
Data Quality Monitoring
02/$ - see front matter & 2009 Elsevier B.V. A
016/j.nima.2009.06.109
esponding author.
ail address: [email protected] (M. De Matti
n behalf of the CMS Collaboration
a b s t r a c t
The size and complexity of the CMS detector makes the Data Quality Monitoring (DQM) system very
challenging. Given the high granularity of the CMS sub-detectors, several approaches and tools have
been developed to monitor the detector performance closely. We describe here the History DQM, a tool
allowing the detector performance monitoring over time.
& 2009 Elsevier B.V. All rights reserved.
1. Introduction
Data Quality Monitoring (DQM) is critically important for theefficient operation of the detector and for the reliable certificationof the recorded data for physics analysis. The CMS experiment atLHC [1] has developed a comprehensive DQM system [2]comprising tools for the creation, filling, transport and archivalof monitoring elements as well as for visualization and retrieval ofthe information. The History DQM takes care of the extraction andvisualization of the summary information obtained from the run-based DQM histograms. The flexible and compact way ofvisualizing the stored information proved to be useful to assessthe data quality during the cosmic data taking of CMS in autumnof 2008.
2. Architecture and design of the History DQM tool
The History DQM (HDQM) consists of three steps:
1.
Extraction and storage of the relevant information in the conditiondatabase: For each histogram corresponding to a selectedquantity the derived summary values are extracted and storedin the CMS database. The list of selected quantities is flexibleand adapted to every DQM task.
2.
Creation and visualization of the trend charts: Two complemen-tary approaches have been developed for the access ofsummary information.(a) A Root based approach: intended for experts to be able toperform detailed analyses.(b) A web approach: intended for prompt feedback.
ll rights reserved.
a).
Several default quantities are available, such as Gaussian fitmean and sigma or landau fit peak and width. It is also possible toprovide user-defined fit functions. Fig. 1 shows a scheme of theHDQM workflow.
2.1. Database container structure
The interaction of the History DQM with the CMS database isbased on the POOL-ORA technology [3], already adopted in theCMS offline software [4] to access the calibration data. Thistechnology is C++ oriented in the definition of the database tableand schemas.
The database schema has been designed to accept and store aconfigurable number of summary information for a configurableand dynamic set of detector elements. This approach allows tostore data at different granularity levels and guarantees thebackward compatibility in case of an extension of the list ofmonitored quantities.
2.2. Information retrieval
There are two approaches available to retrieve the HistoryDQM information.
The web History Data Quality Monitoring service allows to geta quick feedback of the database content, exposing through a webGUI the data trends. It is based on CherryPy [5] and usesMatPlotLib [6] to produce the histograms. CherryPy executespython code which is then linked by boost-python bindings to theCMS software (CMSSW). The user accesses a web interface whichguides him providing hints for the input fields. The web interfaceis shown in Fig. 2. Fig. 3 shows an example of a trend chartobtained with the web interface. Detailed information isaccessible by hovering the mouse pointer on the data points.
ARTICLE IN PRESS
Fig. 1. HDQM workflow scheme.
Fig. 2. Web interface of the HDQM service. It shows the possible options for the various fields. It also allows to blacklist a set of runs and apply preselection cuts.
historicFromT0_V9_ReReco@Tracker@Chi2_CKFTk@mean
Occur: 59.66, IOV: 66989
86.901
69.52
52.14
Occ
uren
cy
34.76
17.38
066467 66569 66657 66711 66878 66952 66987 67038 67544 68087
Fig. 3. Trend plot of the mean number of tracks versus run number produced with the web HDQM service.
D. Giordano et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 263–265264
The Root based inspection of the HDQM database for the trendplot creation is designed for an expert usage. A Root macrointerface allows to query the database and extract trends andcorrelations in a tree-like approach.
Fig. 4. Mean number of reconstructed cosmic tracks per event versus run.
3. Cosmic data taking preliminary results
The Tracker detector has been included in the CMS data takingperiod with cosmic ray trigger in autumn 2008. Data wererecorded both with and without 3.8 T magnetic field, for a total of6 million tracks detected in the CMS Silicon Strip Tracker alongfour weeks of operation. The detector has been constantlymonitored with the DQM tools, inspecting the status and
ARTICLE IN PRESS
Fig. 5. Mean number of reconstructed clusters per track versus run number.
D. Giordano et al. / Nuclear Instruments and Methods in Physics Research A 617 (2010) 263–265 265
performance at different level of granularity, down to the modulelevel (15148 modules). The HDQM has deeply contributed to thecharacterization of the detector over time. The main quantitiesrelated to the Silicon Strip Tracker (number of reconstructedtracks and hits, signal over noise of the reconstructed signal, etc.)have been automatically extracted from the correspondinghistograms and stored in the HDQM database. Figs. 4 and 5show examples of the trends on monitored quantities.
Several analyses were performed to tag the (good/bad) statusof the registered runs. The Root based interface was exploited tocorrelate the stored information in the database and to identifythe properties of the runs deviating from the reference.
4. Conclusions
We have developed a generic tool allowing to follow the timeevolution of the monitoring elements created by the DQM tasks. A
first prototype using the Silicon Strip Tracker data was testedsuccessfully during the cosmic data taking of CMS in autumn of2008. The History DQM proved to be a useful tool to assess thedata quality, thanks to the flexible and compact way of visualizingthe stored information.
References
[1] CMS Collaboration, CERN/LHCC 94-38, Technical proposal, Geneva,Switzerland, 1994.
[2] C. Lenidopoulos, E. Meschi, I. Segoni, G. Eulisse, D. Tsirigkas, Physics and DataQuality Monitoring at CMS, CHEP06, Mumbai, India, February, 2006.
[3] Z. Xie et al., Pool persistency framework for the LHC new developments andCMS applications, in: Proceedings of Frontier Science 2005: New Frontiers inSub Nuclear Physics, Milan, Italy, September 12–17, 2005.
[4] C.D. Jones et al., Analysis environments for CMS, in: Journal of Physics:Conference 2008 Series, vol. 119, 2008, p. 032027.
[5] /http://www.cherrypy.org/S.[6] /http://matplotlib.sourceforge.net/S.