CMS (LHC@CERN) Computing

Total Page:16

File Type:pdf, Size:1020Kb

CMS (LHC@CERN) Computing A common connection: CMS (LHC@CERN) Computing Kajari Mazumdar Department of High Energy Physics Tata Institute of Fundamental Research Mumbai Accelerating International Collaboration and Science through Connective Computation University of Chicago Centre, Delhi March 10, 2015 Introduction • Raison de etre of LHC: discover/rule out existence of Higgs boson particle (if it existed any time in nature?) • Yes, indeed! about 1 pico-second after the Big Bang. • Today, after 13.7 billion years later, CERN-LHC recreates the condition of the very early universe. • Higgs boson discovered within 3 years of start of data taking at LHC. • Experiments study the aftermath of violent collisions of protons/ions -- using very complicated, mammoth detectors -- with ~ 1 million electronic channels taking data every 25 nano-sec. digital summary of information recorded as collision event. • 300 publications per experiment with data collected in Run1 (~4 years) • LHC Computing Grid is the backbone of the success of LHC project. March 10, 2015 K. Mazumdar 2 CERN-LHC 2009-2035: Physics exploitation Nobel prize in 2013, …..? Higgs boson discovery at CERN-LHC in 2012 March 10, 2015 K. Mazumdar 3 What happens in LHC experiment Selection only 1 in 1013 events. we do not know which one! need to check with all! Enormous computing resources needed. March 10, 2015 K. Mazumdar 4 In hard numbers: • 2 big experiments: ATLAS & CMS. • LHC collides 6-8 hundred million protons/second for several years. • Digital information of each collision : 1 – 10 MB • Capacity for Information storage: ~ 1000 collisions /sec. • Physicists must sift through ~ 20 Petabytes (1015 ) of data annually. Higgs search: ~ looking for a particular person in a thousand world populations of today. March 10, 2015 K. Mazumdar 5 Large Collaborative Research • Distributed analysis/computing is essential and designed accordingly. • Made possible by sharing resources across the globe: Crosses organisational, national and international boundaries. • High Energy Physics (HEP) is often compute as well as data intensive. • Experiments performed by several thousand scientists over several years. • CERN-LHC project is an excellent example of large scale collaboration, dealing with huge amount of data. • WWW was born in early 1989 to satisfy the needs of previous generation of high energy experiments carried out by scientists across the world : sharing information. Fine example of application in societal benefit from fundamental research. Becoming necessary in various other fields of research. March 10, 2015 K. Mazumdar 6 From Web to Grid Computing World-wide LHC Computing GRID (conceptualized in ~2000) natural evolution of internet technology in data analysis. 1. Share more than information. 2. Provide resources and services to store/serve ~ 20 PB data/year 3. Provide access to all interesting physics events to ~ 4000 collaborators/expt. Solution through WLCG : Efficient use of resources at many institutes. • Minimize constraints due to user localisation and resource variety • Decentralize control and costs of computing infrastructure much faster delivery of physics! March 10, 2015 K. Mazumdar 7 LHC Computing Grid • Key for grid computing: LAN speed is comparable to the speed of computing processors. • Success due to high speed link across countries. India also pays dividends. • Contributes to few percent level to the computing efforrts of CMS experiment • Typical numbers in LHC for each experiment: # grid WLCG users : ~1000/week, 1M jobs/week # darasets available for analysis across grid: 50k • Today ~ 200 sites across the world active 24X7 for LHC experiments. • ~100 PB disk • Increasing bandwidth fast evolution of LHC computing model in terms of management and processing of data volume From hierarchical structure to parallel/horizontal connections. March opportunistic 10, 2015 usage of resources K. Mazumdar 8 Initial structure of LHC GRID connecting computers across globe Several Petabytes/sec. Online data Tier 0 CERN Experimental recording computer site centre, 10 Gbps Geneva Tier 1 ASIA Germany National centres CERN USA (Taiwan) Italy France Tier 2 1-2.5 Gbps Regional groups in India China Pakistan a continent/nation Korea Taiwan Indiacms T2_IN_TIFR BARC Delhi Panjab Univ. Different Universities, TIFR Univ. Institutes in a country Individual scientist’s PC,laptop, .. March 10, 2015 K. Mazumdar 9 Example: TIER-3 Center for CMS Experiment @ University of Delhi Caters to the needs of the High Energy Physics Group of the Physics Department of the University of Delhi Allows one to fetch CMS data from Tier-2 centres, process data locally, generate monte carlo etc. Kirti Ranjan Tier-3@DU: Hardware Resources Master Server A typical situation Specifications: HP DL180 G6 Processor: 2xIntel Xeon Quad Core (64 bit) Kirti Ranjan Memory : 32GB each Hard disk: 146 GB Worker Node (Total 3 Server) Specifications:HP DL160 G6 Processor: 4xIntel Xeon Quad Core (64 bit) Memory : 16GB each Hard disk: 146 GB Specifications:HP DL160 G8 (New Node Added) Processor: 2xIntel Xeon Quad Core (64 bit) Memory : 16GB each Hard disk: 500 GB One KVM Switch, Console and 1GB Ethernet Switch StorageDisk Space : 12 TB Storage Area 24TB Storage (New Storage Added) Power BackUp: 2*6KV UPS Cooling : AC: 2* 2 Tons each Behind the scene Tier 0: 1M jobs/day. Peak data-transfer rates: 10 gigabytes per second (2 full DVDs of data/s). Users don’t bother about the source of computing plug your computer in to the Internet, it will get the computer power you need to do the job! The Grid infrastructure links together computing resources such as PCs, workstations, servers, storage elements, and provides the mechanism needed to access them. The pervasive nature of grid simply access the Grid through web browser. The Grid is a utility: you ask for computer power or storage capacity and you get it. Middleware is the technical "glue" that allows different computers to "stick" together. We have now experts here offering us to make computing from India easy. March 10, 2015 K. Mazumdar 12 Examples for application of Grid technology WISDOM : search for a cure for malaria MammoGrid : grid for hospitals to share and analyse AstroGrid: astronomy mammograms to improve breast cancer treatment. BIRN: human disease CaBIG: cancer MathCell : grid-managed multi-purpose environment neuGRID: Neuroscience for further research in biology and bioinformatics outGRID: Neuroscience P12S2: learn more about the spread of plant diseases Climateprediction.net: Climate research Virolab: infectious disease Compute Against Cancer: Cancer research ...@home: volunteer computing FightAIDS@Home: HIV/AIDS research FusionGrid: fusion energy Folding@Home: Disease research NEES : earthquakes LHC@home: High energy physics GridRepublic: Many different research projects GIMPS: Mathematics SETI@home: Extraterrestrial intelligence The Charity Engine World Community Grid: Many different projects all with humanitarian aims March 10, 2015 K. Mazumdar 13 Summary • We, from India, are lucky to be part of LHC family. • Grateful for the support received to place India in LHC Grid Map. • Time for India to exert leverage from LHC experience. • Grid computing is changing the way the world is doing science ( as well as business, entertainment, social science and more.) • Efforts must start now to make it happen in India, first the academic community. • Experts are ready to help us in making the leap. Computing aspect of your research can be made easier/bigger! March 10, 2015 K. Mazumdar 14 Warm welcome to the participants. Many thanks to University of Chicago Center, Delhi. March 10, 2015 K. Mazumdar 15 Middleware projects 3G Bridge ESnet Alchemi GLOBUS BioGrid gLite Condor GRIDBUS DCache GridSphere DOE SciDAC IGE EMI ESnet GLOBUS gLite GRIDBUS March 10, 2015 K. Mazumdar GridSphere IGE LHC Grid Map ~ 200 sites across the world active 24X7 ~ 100 PB disk, 300,000 cores CMS and ALICE Tier2 GRID computing March 10, 2015 centersK. Mazumdar in TIFR (Mumbai) and VECC (Kolkata).17 .
Recommended publications
  • The Social Cloud for Public Eresearch
    The Social Cloud for Public eResearch by Koshy John A thesis submitted to the Victoria University of Wellington in fulfilment of the requirements for the degree of Master of Engineering in Network Engineering. Victoria University of Wellington 2012 Abstract Scientific researchers faced with extremely large computations or the re- quirement of storing vast quantities of data have come to rely on dis- tributed computational models like grid and cloud computing. However, distributed computation is typically complex and expensive. The Social Cloud for Public eResearch aims to provide researchers with a platform to exploit social networks to reach out to users who would otherwise be unlikely to donate computational time for scientific and other research ori- ented projects. This thesis explores the motivations of users to contribute computational time and examines the various ways these motivations can be catered to through established social networks. We specifically look at integrating Facebook and BOINC, and discuss the architecture of the functional system and the novel social engineering algorithms that power it. ii Acknowledgments I would first like to thank my parents, John Koshy and Susan John, for their unwavering love and support in all my endeavours. I would like to thank my supervisor, Kris Bubendorfer, for his valuable guidance and support throughout my thesis. Kyle Chard and Ben Palmer have my thanks for their contributions and feedback in the course of au- thoring the IEEE e-Science paper on the Social Cloud for Public eResearch. Many thanks are also due to Andy Linton for his help with managing the development and test server for the Social Cloud for Public eResearch.
    [Show full text]
  • Deploying and Maintaining a Campus Grid at Clemson University Dru Sepulveda Clemson University, [email protected]
    Clemson University TigerPrints All Theses Theses 8-2009 Deploying and Maintaining a Campus Grid at Clemson University Dru Sepulveda Clemson University, [email protected] Follow this and additional works at: https://tigerprints.clemson.edu/all_theses Part of the Computer Sciences Commons Recommended Citation Sepulveda, Dru, "Deploying and Maintaining a Campus Grid at Clemson University" (2009). All Theses. 662. https://tigerprints.clemson.edu/all_theses/662 This Thesis is brought to you for free and open access by the Theses at TigerPrints. It has been accepted for inclusion in All Theses by an authorized administrator of TigerPrints. For more information, please contact [email protected]. Deploying and Maintaining a Campus Grid at Clemson University A Thesis Presented to the Graduate School of Clemson University In Partial Fulfillment of the Requirements for the Degree Masters Computer Science by Dru Sepulveda May 2009 Accepted by: Sebastien Goasguen, Committee Chair Mark Smotherman Steven Stevenson Abstract Many institutions have all the tools needed to create a local grid that aggregates commodity compute resources into an accessible grid service, while simultaneously maintaining user satisfaction and system security. In this thesis, the author presents a three-tiered strategy used at Clemson University to deploy and maintain a grid infrastructure by making resources available to both local and federated remote users for scientific research. Using this approach virtually no compute cycles are wasted. Usage trends and power consumption statistics collected from the Clemson campus grid are used as a reference for best-practices. The loosely-coupled components that comprise the campus grid work together to form a highly cohesive infrastructure that not only meets the computing needs of local users, but also helps to fill the needs of the scientific community at large.
    [Show full text]
  • Volunteer Computing at CERN
    Volunteer Computing at CERN - sustainability model proposal - Supervisors: Openlab Student: Ben Segal, Daniela Dorneanu, Miika Tuisku, [email protected] Francois Grey version 0.1 September 12, 2011 Contents 1 Executive summary 2 2 Introduction 3 2.1 What is Volunteer Computing? . .3 2.2 What is a Volunteer Cloud? . .3 2.3 Volunteer Cloud characteristics . .4 2.4 Case study objectives . .4 3 Analysis 5 3.1 Surveys and interviews for main stakeholders . .5 3.2 Volunteers Profiles . .5 3.3 Brainstorming . .6 4 Use Cases 7 4.1 Volunteer Cloud for CERN . .8 4.2 Volunteer Cloud for Industry - Openlab Model . 11 4.3 Nonprofit Volunteer Cloud . 13 4.4 Commercial Volunteer Cloud . 17 5 Conclusions and future steps 20 6 Acknowledgements 21 7 Annexes 22 7.1 Annex A - CernVM and CoPilot . 22 7.2 Annex B - Detailed Surveys . 22 7.3 Annex C - Detailed description of The Business Model Canvas . 24 8 References 27 1 1 Executive summary This report provides an analysis and evaluation of the possible sustainability models (business models) for Volunteer Cloud Computing. Currently on one side there are millions of volunteers willing to share their CPU power and on the other side there are scientists who need this CPU power but don't really know how to obtain this from the volunteers. Thus the main purpose of this report is to suggest different institutional arrangements that would lower the entry barrier for scientists to Volunteer Cloud Computing, and in particular present such resources to them as a service that was fully compatible with the way they use their current computing infrastructure.
    [Show full text]
  • BOINC: a Platform for Volunteer Computing
    BOINC: A Platform for Volunteer Computing David P. Anderson University of California, Berkeley Space Sciences Laboratory Berkeley, CA 94720 [email protected] December 9, 2018 Abstract “Volunteer computing” is the use of consumer digital devices for high-throughput scientific computing. It can provide large computing capacity at low cost, but presents challenges due to device heterogeneity, unreliability, and churn. BOINC, a widely-used open-source middleware system for volunteer computing, addresses these challenges. We describe its features, architecture, and implementation. Keywords BOINC, volunteer computing, distributed computing, scientific computing, high-throughput computing 1. Introduction 1.1 Volunteer computing Volunteer computing (VC) is the use of consumer digital devices, such as desktop and laptop computers, tablets, and smartphones, for high-throughput scientific computing. Device owners participate in VC by installing a program that downloads and executes jobs from servers operated by science projects. There are currently about 30 VC projects in many scientific areas and at many institutions. The research enabled by VC has resulted in numerous papers in Nature, Science, PNAS, Physical Review, Proteins, PloS Biology, Bioinformatics, J. of Mol. Biol., J. Chem. Phys, and other top journals [1]. About 700,000 devices are actively participating in VC projects. These devices have about 4 million CPU cores and 560,000 GPUs, and collectively provide an average throughput of 93 PetaFLOPS. The devices are primarily modern, high-end computers: they average 16.5 CPU GigaFLOPS and 11.4 GB of RAM, and most have a GPU capable of general-purpose computing using OpenCL or CUDA. 1 The potential capacity of VC is much larger: there are more than 2 billion consumer desktop and laptop computers [2].
    [Show full text]
  • The 10Th BOINC Workshop
    The 10th BOINC Workshop David P. Anderson Space Sciences Lab University of California, Berkeley 29 Sept. 2014 1985 ● Wisconsin UC Berkeley ● Internet as backplane 1987 ● Marionette 1992 ● Industry 1995 ● David Gedye: SETI@home idea 1998 ● SETI@home development – Eric Korpela – Jeff Cobb – Matt Lebofsky 1999 ● SETI@home launch 2000 ● Infrastructure issues ● United Devices 2001 ● United Devices falling-out 2002 ● ClimatePrediction.net: Myles Allen ● BOINC computing power Scientists Volunteers education/outreach 2002 ● Open source software ● Credit ● Replication and validation ● Client job buffer ● Code signing 2002 ● Hiram Clawson, Eric Heien ● NSF proposal – Mari Maeda, Kevin Thompson ● Visit Climateprediction – Carl Christensen, Tolu Aina ● Derrick Kondo ● Vijay Pande 2003 ● UD lawsuit ● Undergrads, PHP code ● Karl Chen, Mr. Python ● Oct: LIGO, Bruce Allen ● Nov: CERN – Francois Grey, Ben Segal ● Nov: WCG kicks tires 2003 job creation MySQL assimilator server scheduler validator transitioner 2004 ● Rom Walton ● Charlie Fenton 2004 ● Anonymous platform ● Separate GUI ● Cross-project ID and credit ● Preemptive scheduling ● Sticky files ● Upload/download hierarchies ● DB as buffer 2004 ● Predictor@home, Michela Taufer – homogeneous redundancy ● SETI@home: Eric Korpela ● BURP: Janus Kristensen ● Climateprediction.net launch ● LHC@home launch 2004 2004 ● Supercomputer 04 talk ● Matt Blumberg, account manager design 2005 ● Einstein@home – Reinhard Prix, Bernd Machenschalk, Oliver Bock ● Primegrid – Rytis Slatkevičius ● Rosetta@home ● IBM World
    [Show full text]
  • BOINC: a Platform for Volunteer Computing
    J Grid Computing (2020) 18:99–122 https://doi.org/10.1007/s10723-019-09497-9 BOINC: A Platform for Volunteer Computing David P. Anderson Received: 21 December 2018 /Accepted: 20 October 2019 /Published online: 16 November 2019 # The Author(s) 2019 Abstract “Volunteer computing” is the use of consum- Physical Review, Proteins, PloS Biology, Bioinformat- er digital devices for high-throughput scientific comput- ics, J. of Mol. Biol., J. Chem. Phys, and other top ing. It can provide large computing capacity at low cost, journals [1]. but presents challenges due to device heterogeneity, About 700,000 devices are actively participating in unreliability, and churn. BOINC, a widely-used open- VC projects. These devices have about 4 million CPU source middleware system for volunteer computing, cores and 560,000 GPUs, and collectively provide an addresses these challenges. We describe BOINC’sfea- average throughput of 93 PetaFLOPS. The devices are tures, architecture, implementation, and algorithms. primarily modern, high-end computers: they average 16.5 CPU GigaFLOPS and 11.4 GB of RAM, and most Keywords BOINC . Volunteer computing . Distributed have a GPU capable of general-purpose computing computing . Scientific computing . High-throughput using OpenCL or CUDA. computing The potential capacity of VC is much larger: there are more than 2 billion consumer desktop and laptop com- puters [2]. Current models have a peak performance (including GPU) of over 3 TeraFLOPS, giving a total 1 Introduction peak performance of 6000 ExaFLOPS. The capacity of mobile devices is similar: there are about 10 billion 1.1 Volunteer Computing mobile devices, and current mobile processors have on-chip GPUs providing as much as 4 TeraFLOPS [3].
    [Show full text]
  • A Review of Desktop Grid Computing Middlewares on Non-Dedicated Resources
    Journal of Theoretical and Applied Information Technology 31st May 2020. Vol.98. No 10 © 2005 – ongoing JATIT & LLS ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 A REVIEW OF DESKTOP GRID COMPUTING MIDDLEWARES ON NON-DEDICATED RESOURCES 1 MAHATHIR RAHMANY, 2ELANKOVAN A. SUNDARARAJAN, 3ABDULLAH MOHD ZIN 1,2,3Fakulti Teknologi Sains dan Maklumat, Universiti Kebangsaan Malaysia (UKM), Bangi, Selangor, 43600, Malaysia E-mail: [email protected], [email protected], [email protected] ABSTRACT Grid Computing is a distributed computing technology that can provide users with powerful computing and storage capacity. Most of the applications of grid computing are now replaced by the use of cloud computing. However, there is an area where the grid computing technology can still be utilized, that is in the area of desktop grid computing (DGC). With the growth of desktop computer belonging to an organization and the increasing of internet penetration worldwide, desktop grid computing can be a good alternative to provide the cheapest high-power computing to solve computing intensive applications by utilizing non-dedicated resources. DGC can also be utilized if an application involved some sensitive and secured data. Another area where DGC can be useful is to support the concept of volunteer computing which is expected be more popular in the future since devices that can be volunteered are not only limited to desktop computers but also other computing devices such as smart phones. The main objective of this paper is to review some available middlewares for DGC. This paper also discusses the future direction of DGC.
    [Show full text]
  • THEME [INFRA-2012-3.3.] [Coordination Actions, Conferences and Studies Supporting Policy Development, Including International Cooperation, for E-Infrastructures]
    THEME [INFRA-2012-3.3.] [Coordination actions, conferences and studies supporting policy development, including international cooperation, for e-Infrastructures] Grant agreement for: Coordination and support action* Annex I - "Description of Work" Project acronym: IDGF-SP Project full title: " International Desktop Grid Federation – Support Project " Grant agreement no: 312297 Version date: Table of Contents Part A A.1 Project summary ......................................................................................................................................3 A.2 List of beneficiaries ..................................................................................................................................4 A.3 Overall budget breakdown for the project ............................................................................................... 5 Workplan Tables WT1 List of work packages ............................................................................................................................1 WT2 List of deliverables .................................................................................................................................2 WT3 Work package descriptions ................................................................................................................... 5 Work package 1......................................................................................................................................5 Work package 2......................................................................................................................................7
    [Show full text]
  • Awareness Watch™ Newsletter by Marcus P
    Awareness Watch™ Newsletter By Marcus P. Zillman, M.S., A.M.H.A. http://www.AwarenessWatch.com/ V9N7 July 2011 Welcome to the V9N7 July 2011 issue of the Awareness Watch™ Newsletter. This newsletter is available as a complimentary subscription and will be issued monthly. Each newsletter will feature the following: Awareness Watch™ Featured Report Awareness Watch™ Spotters Awareness Watch™ Book/Paper/Article Review Subject Tracer™ Information Blogs I am always open to feedback from readers so please feel free to email with all suggestions, reviews and new resources that you feel would be appropriate for inclusion in an upcoming issue of Awareness Watch™. This is an ongoing work of creativity and you will be observing constant changes, constant updates knowing that “change” is the only thing that will remain constant!! Awareness Watch™ Featured Report This month’s featured report covers the Cloud Computing Resources Primer. Cloud Computing has become one of the more significant topics when it comes to computing and the Internet in the 2010/2011 time period and continues on. Discovering what clouds are, what types of clouds are available and cloud security have become popular quests for knowledge discovery! The primer gives you this information and much more with the many many resources about cloud computing available from the Internet including grids, distributed, and cloud computing. The information has been taken from my constantly updated Subject Tracer™ titled Grid, Distributed and Cloud Computing Resources and available directly from: http://www.GridResources.info/ 1 Awareness Watch V9N7 July 2011 Newsletter http://www.AwarenessWatch.com/ [email protected] eVoice: 800-858-1462 © 2011 Marcus P.
    [Show full text]
  • Volunteer Computing
    Volunteer Computing David P. Anderson Space Sciences Lab U.C. Berkeley May 2, 2007 Outline z Goals of volunteer computing z How BOINC works z Some volunteer computing projects z Some research directions z Non-technical problems Goal: Use all the computers in the world, all the time, to do worthwhile things z What do we mean by “computers”? z Who owns the computers? − Individuals (60% and rising) − Organizations z What does “worthwhile” mean? BOINC (Berkeley Open Infrastructure for Network Computing) z Middleware for volunteer computing z Open-source (LGPL) z Application-driven 60% Accounts PC 40% Attachments with Projects resource share The volunteer computing game Volunteers Projects Internet z Do more science z Involve public in science Volunteer computing != Grid computing Resource anonymous, unaccountable; identified, accountable owners need to check results Managed yes – software stack no – need plug & play software systems? requirements OK Clients behind yes – pull model no – push model firewall? yes ISP bill? no ... nor is it “peer-to-peer computing” Types of utility z Compute cycles (usually FLOPS) − with latency bound − with RAM − with storage − guaranteed availability z Storage − network bandwidth − availability z Network bandwidth at specific time z Wide/diverse deployment z Human steering of computation What is resource share? z Given: set of hosts, attachments z “Schedule”: what projects use resources when z Ideally: − Weighted Pareto optimality of utility? − Across multiple hosts z In BOINC − determines usage of bottleneck
    [Show full text]