Enabling Grids for E-sciencE GridICE: overview and current status Guido Cuscela INFN – Bari Service Challenge Technical Meeting September 15, 2006

Enabling Grids for E-sciencE 2 Outline Old and new features (new release) What we are monitoring Job monitoring results (INFN-T1) Use cases (web interface) Issues

Enabling Grids for E-sciencE 3 Why GridICE for monitoring –Grid monitoring  Grid resources and services are subject to failures and is fundamental their monitoring for the Grid utilization –Local monitoring  GridICE can be used to monitor your own farm (in connection with a local server)

Enabling Grids for E-sciencE 4 Present deployment EGEE EGEE-SWE RDIG EGEE-SEE Grid.it GILDA CMS ATLAS EUMedGrid EUChinaGRID BalticGrid EELA BeGrid Version: v was released on Fri, 08 Sep 2006 – The Grid.it server has been already Installed servers are monitoring Grid resources in the scope of: The EGEE server runs since July 2005 The Grid.it server runs since July 2005 without any major intervention and continue to perform very well

Enabling Grids for E-sciencE 5 How does it work Generation Distribution Presentation Processing Sensors enquiring entities and encoding the measurements according to a schema Transmission of the events from the source to any interested parties Abstract the huge number of received events in order to enable the consumer to draw conclusions about the operation of the monitored system e.g., filtering according to some predefined criteria, or summarising a group of events

Enabling Grids for E-sciencE 6 Features powerful and complete web-based interface for data presentation each view of the web-based interface offers the same data in XML format support for customized graph generation notification service –Customizable monitoring of nodes automatic discovery of new resources to be monitored through the Grid Information Service complete set of monitored metrics, from host-related to Grid service related characteristics –supports and extends the GLUE Schema support for the following batch systems: OpenPBS, Torque, LSF integrated with network-related infrastructure for monitoring the connectivity of a Grid

Enabling Grids for E-sciencE 7 What we are monitoring Hardware monitoring: –fabric level monitoring via LEMON sensors Services monitoring: –For every grid node we check the related services (via standard GRIIS) –Monitoring of every process/daemon which has to run on nodes Job monitoring: –New “lightweight” job monitoring sensors (we are running at INFN-T1 with no problems and with more than 3000 jobs R/Q) –Execution time reduced of the order of a factor ten compared with the previous version –About 99% of jobs retrieved correctly LRMS monitoring (since GridICE release): –LRMSinfo sensor as preliminary SLA support and basic site CPU usage efficiency –No sensors on WNs (all needed information retrieved on the CE from batch system)

Enabling Grids for E-sciencE 8 Fabric monitoring

Enabling Grids for E-sciencE 9 Job monitoring Comparison between BOSS and GridICE jobs data (CMS production aggregate data from INFN-T1,INFN-Legnaro,INFN-Bari,INFN-Pisa) Total number of jobs 5939 (3175 at INFN-T1) Number of jobs not seen by GridICE 97 (55 at INFN-T1) 98.3% accuracy

Enabling Grids for E-sciencE 10 New features in release v1.9.0 Region/ROC support –filter the resource by region –modify site/region binding Synchronization with GOCDB –Detailed info on site downtimes (foreseen, partial or global) LRMSInfo –a bunch of new charts available to have a view of resources utilization More options to retrieve jobs information (search by global-ID, local user …) New statistic plots with new look & feel (ex: Grid Jobs vs. Local Jobs) Chart Section Reorganized –new menu to select single charts or per user role view Clean Up DB History –available a new script that help in deleting historical data from the DB (you should need to delete data older then a specific date/time)

Enabling Grids for E-sciencE 11 Different viewpoints We focus on the following categories of users: – VO manager  actual set of resources accessible to VO members  “How many jobs submitted by my users are running or queued?” – Grid operator  all resources under responsibility of a Grid Operator Center  “How many resources are available?” – Site administrator  site resources offered to a Grid  “Is there any service down?”

Enabling Grids for E-sciencE 12 Host View

Enabling Grids for E-sciencE 13 Host View - Details

Enabling Grids for E-sciencE 14 Job View

Enabling Grids for E-sciencE 15 Local monitoring

Enabling Grids for E-sciencE 16 GOC interfacing

Enabling Grids for E-sciencE 17 LRMSinfo

Enabling Grids for E-sciencE 18 Issues Queries lateness [end of the year] –We are working on database improvements (table partitioning, db schema modification …) LeMON 2.10.x [end of the year] –We have planned to migrate to latest LeMON version as soon as possible gLite 3.0 [end of October] –Integration of job monitoring sensors is finished (we are testing them with italian ROC release team) Storage probes [end of October] –Grid transfer monitoring (DPM, CASTOR, dCache) –local transfer and access to file (RFIO,dcap; both authenticated and un- authenticated versions ) –Not yet ready for production. Need some more development and tests Advanced RB probe –Code is ready for gLite. We need some more time to integrate the info on the GridICE collecting infrastructure FTS monitoring –Used at CNAF –Will be integrated in GridICE Group and VOMS roles monitoring –Will be available in new releases

Enabling Grids for E-sciencE 19 Conclusions We are able to provide a wide and easy to use Grid monitoring –Fabric level –Services monitoring –Job monitoring –Storage and FTS monitoring (shortly) We keep on working to improve: –Performances –Reliability –Design We are open to collect new requirements and support your monitoring needs

Enabling Grids for E-sciencE 20 References GridICE Publications: [1] S. Andreozzi, N. De Bortoli, S. Fantinel, A. Ghiselli, G. L. Rubini, G. Tortone, M. C. Vistoli GridICE: a monitoring service for Grid systems, Future Generation Computer System 21 (2005) 559–571 [2] C. Aiftimiei, S. Andreozzi, G. Cuscela, N. De Bortoli, G. Donvito, S. Fantinel, E. Fattibene, G. Misurelli, A. Pierro, G.L. Rubini, G.Tortone. GridICE: Requirements, Architecture and Experience of a Monitoring Tool for Grid Systems. In Proceedings of the International Conference on Computing in High Energy and Nuclear Physics (CHEP2006), Mumbai, India February [3] C. Aiftimiei, S. Andreozzi, G. Cuscela, N. De Bortoli, G. Donvito, S. Fantinel, E. Fattibene, G. Misurelli, A. Pierro, G.L. Rubini, G.Tortone. Flexible notification service for Grid monitoring events. In Proceedings of the International Conference on Computing in High Energy and Nuclear Physics (CHEP2006), Mumbai, India February [4] S. Andreozzi, A. Ciuffoletti, A. Ghiselli, C. Vistoli. Monitoring the Connectivity of a Grid. In Proceedings of the 2nd International Workshop on Middleware for Grid Computing (MGC 2004) in conjunction with the 5th ACM/IFIP/USENIX International Middleware Conference, Toronto, Canada, October GridICE dissemination:

Enabling Grids for E-sciencE 21 Backup slides

Enabling Grids for E-sciencE 22 VO View Use Case 3 VO manager Detecting all Grid resources for the “alice” VO

Enabling Grids for E-sciencE 23 Job monitoring load JM off JM on

Enabling Grids for E-sciencE 24 New charts selection