LargeScale Computing and Data Management Fons Rademakers CERN

  • Slides: 15
Download presentation
Large-Scale Computing and Data Management Fons Rademakers CERN openlab CTO

Large-Scale Computing and Data Management Fons Rademakers CERN openlab CTO

The Large Hadron Collider (LHC) 20/5/2015 Visit Lithuanian Delegation 2

The Large Hadron Collider (LHC) 20/5/2015 Visit Lithuanian Delegation 2

Big Data ! Detectors: 7000 tons “microscopes” 150 million sensors Select 40 million times

Big Data ! Detectors: 7000 tons “microscopes” 150 million sensors Select 40 million times per second Peta Bytes / sec ! Level-1 Filters (Triggers) Select 100, 000 times per second Tera Bytes / sec ! High-Level Filters Select 100 times per second Giga Bytes / sec ! 20/5/2015 Visit Lithuanian Delegation 3

20/5/2015 Visit Lithuanian Delegation 4

20/5/2015 Visit Lithuanian Delegation 4

Reconstruction Simulation Storage Distribution CPUs Network Databases # of cores: # of disks: #

Reconstruction Simulation Storage Distribution CPUs Network Databases # of cores: # of disks: # of processors: 99, 061 80, 754 18, 633 # of 10 GB NICs: # of 1 GB NICs: # of servers: Disk space (PB): RAM memory (TB): 3, 218 19, 334 10, 809 129 349 Storage Infrastructure 5

The Grid Tier-0 (CERN): • Data recording • Initial data reconstruction • Data distribution

The Grid Tier-0 (CERN): • Data recording • Initial data reconstruction • Data distribution Tier-1 (12 centres): • Permanent storage • Re-processing • Analysis 20/5/2015 Visit Lithuanian Delegation Tier-2 (68 Federations, ~140 centres): • Simulation • End-user analysis • 525, 000 cores • 450 PB 6

CERN Data Centres Dante/Géant 100 Gb/s T-Systems 20/5/2015 Visit Lithuanian Delegation 7

CERN Data Centres Dante/Géant 100 Gb/s T-Systems 20/5/2015 Visit Lithuanian Delegation 7

Evolution of Computing requirements Higher trigger (data) rates driven by physics needs Based on

Evolution of Computing requirements Higher trigger (data) rates driven by physics needs Based on understanding of likely LHC parameters; Foreseen technology evolution (CPU, disk, tape) Experiments work hard to fit within constant budget scenario Estimated evolution of requirements 2015 -2017 2008 -2013: Actual deployed capacity Line: extrapolation of 2008 -2012 actual resources Curves: expected potential growth of technology with a constant budget • CPU: 20% yearly growth • Disk: 15% yearly growth 20/5/2015 Visit Lithuanian Delegation 8

International Collaborations Private Public 20/5/2015 Collaborations Visit Lithuanian Delegation 9

International Collaborations Private Public 20/5/2015 Collaborations Visit Lithuanian Delegation 9

Information Technology Research Areas Data acquisition and filtering Computing platforms, data analysis, simulation Data

Information Technology Research Areas Data acquisition and filtering Computing platforms, data analysis, simulation Data storage and long-term data preservation Compute provisioning (cloud) Networks Medical applications 20/5/2015 Visit Lithuanian Delegation Data analytics 10

New professional profiles Multicore CPU programming, graphical processors (GPU), multithreaded software Data analysis platforms,

New professional profiles Multicore CPU programming, graphical processors (GPU), multithreaded software Data analysis platforms, statistics, mathematics, data visualization, monitoring, security, etc. Software & Computing Engineers Data Scientists 20/5/2015 Visit Lithuanian Delegation Applications of physics to medical research (simulation, data analysis, image processing) Multidisciplinary applications 11

Activities § Identification of projects and IT services • Many open questions: understand computing/data

Activities § Identification of projects and IT services • Many open questions: understand computing/data models, • governance and handling policies, offline vs. real-time, evolution, code modernization, etc. Collaboration between CERN and other institutes, possibly industrial companies as needed/desirable § IT in Healthcare Workshop • Co-organized by CERN openlab and CERN Medical Applications Group 20/5/2015 Visit Lithuanian Delegation 12

Proposed initial services Digital content management system Publications, datasets, images, software Community-curated content, backed

Proposed initial services Digital content management system Publications, datasets, images, software Community-curated content, backed by CERN Data Storage facilities and policies. Largest research Open. Stack private cloud infrastructure Professionally managed across the two CERN data centres A programmable, extensible, scaleable platform for data analysis, storage, sharing, reproducibility and long-term preservation. Applications as a service. 20/5/2015 Visit Lithuanian Delegation 13

Code Modernization Project § 20/5/2015 Code modernization for leading edge cell development code, Cx

Code Modernization Project § 20/5/2015 Code modernization for leading edge cell development code, Cx 3 D, move code from Java to C++ and optimize for modern multi-core vector architectures and co-processors and GPU’s Visit Lithuanian Delegation 14

EXECUTIVE CONTACT Alberto Di Meglio, CERN openlab Head (Phase V) alberto. di. meglio@cern. ch

EXECUTIVE CONTACT Alberto Di Meglio, CERN openlab Head (Phase V) alberto. di. meglio@cern. ch TECHNICAL CONTACT Fons Rademakers, CERN openlab CTO fons. rademakers@cern. ch COMMUNICATION CONTACT Mélissa Gaillard, CERN openlab Communication Officer melissa. gaillard@cern. ch ADMIN CONTACT Kristina Gunne, CERN openlab Administration Officer kristina. gunne@cern. ch 21/11/2014 Visit Lithuanian Delegation 15