RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator

Page created by Zachary Bailey
 
CONTINUE READING
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES updates, resources and Access
    European HPC ecosystem

                                  Sergi Girona
                               RES Coordinator
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES: HPC Services for Spain

      •The RES was created in 2006.
      •It is coordinated by the Barcelona Supercomputing
      Center (BSC-CNS).
      •It forms part of the Spanish “Map of Unique Scientific
      and Technical Infrastructures” (ICTS).
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES: HPC Services for Spain
RES is made up of 12 institutions and 13 supercomputers.

                  BSC MareNostrum
                  CESGA FinisTerrae
             CSUC Pirineus & Canigo
                     BSC MinoTauro
                           UV Tirant
                        UC Altamira
                       UPM Magerit
     CénitS Lusitania & SandyBridge
                        UMA Picaso
                       IAC La Palma
                  UZ CaesarAugusta
                   SCAYLE Caléndula
                       UAM Cibeles
                                       1           10                    100   1000
                                           TFlop/s (logarithmic scale)
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES: HPC Services for Spain
• Objective: coordinate and manage high performance computing services to promote
  the progress of excellent science and innovation in Spain.
• It offers HPC services for non-profit, open R&D purposes.
• Since 2006, it has granted more than 1,000 Million CPU hours to 2,473 research
  activities.
                                                                Research areas

 Hours granted per area

                             23%
    AECT        30%                             Mathematics, physics     Astronomy, space
                                                 and engineering         and earth sciences
    BCV
    FI
    QCM                        19%

                      28%

                                                  Life and health         Chemistry and
                                                      sciences           materials sciences
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES supercomputers
BSC (MareNostrum 4) 165888 cores, 11400 Tflops
Main processors:   Intel(R) Xeon(R) Platinum 8160
Memory:            390 TB
Disk:              14 PB

UPM (Magerit II)    3920 cores, 103 Tflops
Main processors :   IBM Power7 3.3 GHz
Memory:             7840 GB
Disk:               1728 TB

UMA (Picasso)       4016 cores, 84Tflops
Main processors:    Intel SandyBridge-EP E5-2670
Memory:             22400 GB
Disk:               720 TB

UV (Tirant 3)       5376 cores, 111,8 Tflops
Main processors:    Intel SandyBridge-EP E5-2670
Memory:             10752 GB
Disk:               14 + 10 TB

CSUC (Pirineus)     2784 cores, 283,66 Tflops
Main processors:    Intel(R) Xeon(R) Platinum 8160
Memory:             12000 GB
Disk:               200 TB

CSUC (Canigo)       384 cores, 33,2 Tflops
Main processors:    Intel(R) Xeon(R) Platinum 8160
Memory:             9000 GB
Disk:               200 TB
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES supercomputers
CénitS (Lusitania 2)   800 cores, 33,2 Tflops
Main processors        Intel Xeon E5-2660v3, 2.6GHz
Memory:                10 GB
Disk:                  328 TB

CénitS (SandyBridge)   2688 cores, 56 Tflops
Main processors        Intel Sandybridge Xeon
Memory:                5376 GB
Disk:                  328 TB

BSC (MinoTauro)        624 cores, 251 Tflops
Main processor:        39x 2 Intel Xeon E5-2630 v3
Memory:                20 TB
Disk:                  14PB (shared with MN4)

CESGA (FinisTerrae 2) 7712 cores, 328,3Tflops
Main processor:       Intel Xeon E5-2680v3
Memory:               40 TB
Disk:                 960 TB

UC (Altamira 2+)       5120 cores, 105 Tflops
Main processor:        Intel SandyBridge
Memory:                15,4 TB
Disk:                  2PB

UZ (Caesaraugusta)     2014 cores, 80.5 Tflops
Main processor:        Intel E5-2680v3, 2.5GHz
Memory:                5400 GB RAM memory
Disk:                  219TB
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
RES supercomputers
SCAYLE (Caléndula)   2432 cores, 50,6 Tflops
Main processor:      Intel SandyBridge Xeon
Memory:              4864 GB
Disk:                600 TB

UAM (Cibeles)        368 cores, 14,1 Tflops
Main processor:      Intel Xeon E5-2630 v3, 2.40GHz
Memory:              896 GB
Disk:                80 TB

UAM (SandyBridge) – coming soon – 2688 cores, 56Tflops
Main processor:     Intel SandyBridge Xeon, 2.60GHz
Memory:             5376 GB
Disk:               80 TB

IAC (LaPalma)        4032 cores, 83,85 Tflops
Main processor:      Intel SandyBrigde
Memory:              8064 GB
Disk:                60 TB
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
Resources granted: CPU hours
               400.000

               350.000

               300.000

                                                         140mh available
               250.000
Hours x 1000

               200.000

               150.000

               100.000

                50.000

                    0
                         2006   2007   2008   2009      2010     2011   2012   2013      2014        2015   2016   2017

                                                     Requested hours           Awarded hours (A+B)
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
How to apply?
•   RES resources are open to researchers and spin-offs:

     o Computing resources: CPU hours and storage
     o Technical support: application analysis, porting of applications, search for the
       best algorithm… to improve performance and ensure the most effective use of
       HPC resources.
     o Free of cost at the point of usage

                                                                         Nex
                                                                             td
•   Three open competitive calls per year.                               Janu eadline
                                                                             ary
                                                                                 2019 :

                                      Deadline for
                   Period                                  Starting date
                                      applications
                     P1                 January              1st March
                     P2                   May                 1st July
                     P3               September            1st November
RES updates, resources and Access European HPC ecosystem - Sergi Girona RES Coordinator
How to apply?
                     RES intranet: https://www.bsc.es/res-intranet

•   Researchers present a proposal which includes research project description,
    technical requirements and research group experience.

•   Accepted proposals have access to RES supercomputers for 4 months.

•   Granted time can be: hours with priority (hours A) or without priority (hours B)
Proposal evaluation

                           Formal            Technical
                          evaluation         evaluation
Submit

         Access committee              Scientific experts panel

           Final report of
         accepted activities
Activity length
   • Accepted proposals have access to RES supercomputers for 4 months. If
     your activity needs more time to be properly developed, you can ask for a
     continuation activity:

                          New activity              Report dissemination
                          (4 months)                    information

                    Continuation activity           Report dissemination
                        (4 months)                      information

                             • The application form is simplified
                             • Are preferably allocated to the same machine
Continuation activities
                             • In the evaluation, one reviewer is kept from the
                               previous activity and the second reviewer changes
RES Users’ Committee
•   CURES aims to provide advice and feedback to RES coordinators:

     o Promotes optimal use of high performance computing facilities

     o Shares information about users’ experiences

     o Voices user concerns

•   You can contact CURES through RES intranet:
Which is/could be the main impediment to
         apply to RES resources?

 Ø I don't know how to write a strong application
 Ø I'm not sure if I can apply
 Ø Lack of HPC expertise in my research group
 Ø Too much paperwork
 Ø Not enough resources for my project
Tips to write a strong proposal

•   Read carefully all the protocols, guides and FAQs in:
                         https://www.res.es/en/access-to-res

•   Project description section: highlight the importance of your project, not
    only from the scientific point of view but also from the return to society.
     Ø Why do you think that your project should deserve the resources requested?
Tips to write a strong proposal
•   Activity description section: specify clearly why do you need
    supercomputing resources. Write as accurate as possible the flowchart of
    the simulations. Indicate that you have your own human resources in your
    group to run and process the output of all the simulations you propose.
     Ø Why do you need to carry out the simulations in the selected machine?

     Ø Is the amount of computing resources requested adjusted to your needs and
        properly justified?

•   Doubts about software/HPC resources: ask support team!
     Ø Are your jobs adequate for parallel computing?

                              support@bsc.es
Which is/could be the main impediment to
         apply to RES resources?

 Ø I don't know how to write a strong application
 Ø I'm not sure if I can apply
 Ø Lack of HPC expertise in my research group
 Ø Too much paperwork
 Ø Not enough resources for my project
Who can apply?

• RES resources are aimed at open R+D+I activities:

   o Researchers from academia and public R&D institutions
   o Spin-offs during their first 3 years from its creation
   o Collaboration projects between private companies and research groups
      from academia or public institutions
   o Open to international applicants, but we recommend the collaboration
      with researchers from Spanish institutions
Which is/could be the main impediment to
         apply to RES resources?

 Ø I don't know how to write a strong application
 Ø I'm not sure if I can apply
 Ø Lack of HPC expertise in my research group
 Ø Too much paperwork
 Ø Not enough resources for my project
RES events: technical training
These workshops are organized by the RES nodes and aim to provide the
knowledge and skills needed to use and manage the supercomputing
facilities.

• Check the agenda in RES website:
  https://www.res.es/en/events?event_type=technical_training

• PATC courses in BSC (PRACE Advanced Training Center):
  https://www.bsc.es/education/training/patc-courses
RES events: networking opportunities
Scientific seminars
The RES promotes scientific seminars which address supercomputing
technology applications in specific scientific areas. These events are mainly
organized by RES users and are open to the entire research community.

            In 2017:

        ü 5 scientific seminars
        ü More than 300 attendees

      Sep       Next Generation Sequencing and Supercomputing: life as a couple

      27               CBMSO-UAM (Madrid)

                         Agenda 2018: www.res.es/en/events
RES events: networking opportunities

 RES Users’ Meeting: 20 September 2018 - Valencia

The agenda includes:
    •   Information about RES and the European HPC ecosystem
    •   Plenary session: Research Open Data
    •   Parallel scientific sessions
    •   Poster session
    •   Networking opportunities
    •   Evening social event
               www.res.es/users-conference-2018
Funded by the EC: 2017 - 2021

ü Mobility grants for researchers using HPC resources
ü Short stays to visit scientific hosts (3 weeks – 3 months)
ü Funds for travel and living allowance
ü Access to European HPC facilities

                     Nex
                         td
                     20 S eadline
                         epte     :
                             mbe
                                r

      http://www.hpc-europa.eu/
Which is/could be the main impediment to
         apply to RES resources?

 Ø I don't know how to write a strong application
 Ø I'm not sure if I can apply
 Ø Lack of HPC expertise in my research group
 Ø Too much paperwork
 Ø Not enough resources for my project
RES forms
In the RES we try to keep the administrative procedures short and simple for
researchers:

     • New activity application form: 10 pages on average

     • Continuation activity application form: simplified

     • Dissemination form: 3 pages on average
          o   Brief description of results (1-2 paragraphs)
          o   Publications
          o   Figures / pictures
          o   Optional: patents, PhD students…

     • Intermediate reports: 1-2 sentences (“Everything is ok”)

     • Resubmission of non-accepted activities: one click
Which is/could be the main impediment to
         apply to RES resources?

 Ø I don't know how to write a strong application
 Ø I'm not sure if I can apply
 Ø Lack of HPC expertise in my research group
 Ø Too much paperwork
 Ø Not enough resources for my project
PRACE HPC Access
• Call for Proposals for Project Access:
                                                     Nex
                                                         t de
                                                      30 O adline:
    - 12, 24 or 36-month projects                          ctob
                                                               er
    - Minimum request: 30 million core hours

• Call for Proposals for PRACE Preparatory Access:

    - From 2 to 12 month projects

                       http://www.prace-ri.eu/
EuroHPC
Distributed Supercomputing Infrastructure

24 members, including
5 Hosting Members
(Switzerland, France, Germany,                                      Hazel Hen
Italy and Spain)                                       JUQUEEN

                                                                                SuperMUC
524 scientific projects
enabled                                   Curie

70 PFlops/s of peak performance                                   Piz Daint
on 7 world-class systems
                                                                    Marconi

>10.000 people trained by 6 PRACE
                                                  MareNostrum
Advanced Training Centers and
others events

                                 Access   prace-ri.eu/hpc_acces
Top500 (June 2018), European ranking

                                                                                                                Cores &      Rmax        Rpeak
Europe   Rank               Name                     Computer                          Site           Country
                                                                                                                 Accel.    [PFlop/s]    [PFlop/s]
                                         Cray XC50, Xeon E5-2690v3 12C
                                                                                                                  361760
  1       6     Piz Daint                2.6GHz, Aries interconnect , NVIDIA   CSCS           Switzerland                       19,59        25,33
                                                                                                                  297920
                                         Tesla P100
                                         Proliant DL380 Gen10, Xeon Platinum
                                                                                                                  253600
  2       13    HPC4                     8160 24C 2.1GHz, Mellanox             ENI            Italy                             12,21        18,62
                                                                                                                  177520
                                         InfiniBand EDR, NVIDIA Tesla P100
                                         Bull Sequana X1000, Intel Xeon Phi
  3       14    Tera-1000-2                                                    CEA            France              561408        11,97        23,40
                                         7250 68C 1.4GHz, Bull BXI 1.2
                                         CINECA Cluster, Lenovo
                                         SD530/S720AP, Intel Xeon Phi 7250
  4       18    Marconi Intel Xeon Phi                                         CINECA         Italy               312936         8,41        16,21
                                         68C 1.4GHz/Platinum 8160, Intel
                                         Omni-Path
                                         Cray XC40, Xeon E5-2695v4 18C
  5       20                                                                   UKMET          United Kingdom      241920         7,04         8,13
                                         2.1GHz, Aries interconnect
                                         Lenovo SD530, Xeon Platinum 8160
  6       22    MareNostrum                                                    BSC            Spain               153216         6,47        10,30
                                         24C 2.1GHz, Intel Omni-Path
                                         Bull Sequana X1000, Xeon Platinum
                                         8168 24C 2.7GHz, Mellanox EDR
  7       23    JUWELS Module 1                                                Juelich        Germany             114480         6,18         9,89
                                         InfiniBand/ParTec ParaStation
                                         ClusterSuite
                                         Cray XC40, Xeon E5-2680v3 12C
  8       27    Hazel Hen                                                      HLRS           Germany             185088         5,64         7,40
                                         2.5GHz, Aries interconnect
                                         Intel Compute Module HNS2600BP,       Max-Planck-
  9       28    COBRA                    Xeon Gold 6148 20C 2.4GHz, Intel      Gesellschaft   Germany             127520         5,61         9,79
                                         Omni-Path                             MPI/IPP
                                         SGI ICE X, Xeon Xeon E5-2670/ E5-
  10      30    Pangea                                                         Total          France              220800         5,28         6,71
                                         2680v3 12C 2.5GHz, Infiniband FDR
Top10 industrial machines (June 2018)

                                                                                                                             Cores &   Rmax        Rpeak
Rank          Name                               Computer                                      Site            Country
                                                                                                                              Acc.   [PFlop/s]    [PFlop/s]
                      Proliant DL380 Gen10, Xeon Platinum 8160 24C 2.1GHz, Mellanox                                            253600
 13    HPC4                                                                             Eni S.p.A.           Italy                        12,21        18,62
                      InfiniBand EDR, NVIDIA Tesla P100                                                                        177520
                      SGI ICE X, Xeon Xeon E5-2670/ E5-2680v3 12C 2.5GHz, Infiniband    Total Exploration
 30    Pangea                                                                                                France            220800      5,28          6,71
                      FDR                                                               Production
                                                                                        Petroleum Geo-
 35    Abel           Cray XC30, Xeon E5-2698v3 16C 2.3GHz, Aries interconnect                               United States     145920      4,04          5,37
                                                                                        Services
                      NVIDIA DGX-1/Relion 2904GT, Xeon E5-2698v4 20C 2.2GHz/ E5-                                                60512
 45                                                                                     Facebook             United States                 3,31          4,90
                      2650v4, Infiniband EDR, NVIDIA Tesla P100/Quadro GP100                                                    55552
                      NVIDIA DGX-1, Xeon E5-2698v4 20C 2.2GHz, Infiniband EDR, NVIDIA                                           60512
 46    DGX Saturn V                                                                   NVIDIA Corporation United States                     3,31          4,90
                      Tesla P100                                                                                                55552
                      iDataPlex DX360M4, Intel Xeon E5-2680v2 10C 2.8GHz, Infiniband                                            72000
 49    HPC2                                                                             Eni S.p.A.           Italy                         3,19          4,61
                      FDR, NVIDIA K20x                                                                                          42000
                      Lenovo NeXtScale nx360M5, Xeon E5-2697v4 18C 2.3GHz, Infiniband   Energy Company                          66000
 65    HPC3                                                                                                  Italy                         2,59          3,80
                      EDR, NVIDIA Tesla K80                                             (A)                                     39000
                      PowerEdge R440/C6320p, Intel Xeon Phi 7250/Xeon Gold 6130 16C
 78    Makman-3                                                                         Saudi Aramco         Saudi Arabia       53300      2,32          3,58
                      2.1GHz, Intel Omni-Path
                      Inspur TS10000, Xeon Gold 6130 16C 2.1GHz, NVIDIA Tesla V100,                                             55104
 79                                                                                     Internet Service P   China                         2,29          4,89
                      25G Ethernet                                                                                              52480

 80    Makman-2       Dell PowerEdge R630, Xeon E5-2680v3 12C 2.5GHz, Infiniband QDR    Saudi Aramco         Saudi Arabia       76032      2,25          3,04
Status of Signatory Countries

                          Countries
                          Signatory European
EuroHPC mission and objectives

• To develop, deploy, extend and maintain in the Union an
  integrated world-class supercomputing and data infrastructure
  and to develop and support a highly competitive and innovative
  High-Performance Computing ecosystem

                                                                                        R&I, Applications
                                                                     Infrastructure &
• Provide European scientists, industry and the public sector with
  the latest HPC and data infrastructure and support the

                                                                        Operations
  development of its technologies and applications across a wide

                                                                                             & Skills
  range of fields
• Support an ambitious research and innovation agenda to develop
  and maintain in the Union world-class High Performance
  Computing ecosystem, exascale and beyond, covering all
  scientific and industrial value chain segments, including low-
  power processor and middleware technologies, algorithms and
  code design, applications and systems, services and engineering,
  interconnections, knowhow and skills, for the next generation         HPC Ecosystem
  supercomputing era
Infrastructure and Operation

• Two exascale systems
   • Possibly one on 2022 and a second one in 2023
   • At least one with European competitive technologies
   • Expected budget on TCO: 500 m€ each
• Two pre-exascale systems
   • In operation in January 2021
   • Expected budget on TCO: 240m€ each
• At least 2 petascale systems
   • In operation in January 2021
   • Budget: TBD
• Next steps
   • Identify the hosting sites
   • User requirements to prepare RFP
EuroHPC Roadmap

HPC and Accelerator Chips &
Systems                                                         Gen 1             Gen 2
           2018           2019                      2020          2021     2022     2023       2024           2025
                                EuroHPC JU
                                                                                                            Software
                                   App & Software
                                                     Software

                                                     sw stack

                                                     Apps                                                Applications
        R&I

                                                      Co-Design

                                                    EsD            Intg.                                  Integration

                              Exascale technologies integration

     Procurement                                    2 Pre-Exa                        2 Exascale (1 EU)
                                                     2 Peta
Visit our website: www.res.es

              Subscribe to our newsletter

Contact us!
              Follow us in Twitter:
              @RES_HPC

              applications@res.es
              dissemination@res.es
THANK YOU!
You can also read