Report from the Scientific Program Committee Anna Hasenfratz Task - - PowerPoint PPT Presentation

report from the scientific program committee
SMART_READER_LITE
LIVE PREVIEW

Report from the Scientific Program Committee Anna Hasenfratz Task - - PowerPoint PPT Presentation

USQCD All-Hands Meeting 2017 Report from the Scientific Program Committee Anna Hasenfratz Task of the SPC Follow the USQCD Charter: The Scientific Program Committee (SPC) advises the Executive Committee (EC) - The SPC advises the EC on


slide-1
SLIDE 1

USQCD All-Hands Meeting 2017

Report from the Scientific Program Committee

Anna Hasenfratz

slide-2
SLIDE 2

Task of the SPC

Follow the USQCD Charter: The Scientific Program Committee (SPC) advises the
 Executive Committee (EC)

  • The SPC advises the EC on science priorities for

USQCD

  • The SPC suggests to the EC allocations of computer

time on the USQCD facilities and zero-priority time on Incite resources

slide-3
SLIDE 3

Scientific Program Committee

  • Tom Blum
  • Aida El-Khadra (Co-chair, incoming chair)
  • Steve Gottlieb
  • Anna Hasenfratz (Chair)
  • Keh-Fei Liu (replaces Will Detmold)
  • David Richards (replaces Kostas Orginos)
  • Swagato Mukherjee
slide-4
SLIDE 4

Available resources: (from CfP)

– BNL BG/Q: 17.7M BG/Q core-hrs (29M Jpsi core-hrs)

(last year: 71M - this resource is going away) – Clusters: 261M (last year: 263M) Jpsi core-hrs 71M @ JLAB, 190M at FNAL – GPUs: 8.4M (last year: 7.7M) Fermi/Tesla GPU-hrs (688 M Jpsi core-hrs): 2.6+3.4+2.4 @ JLAB, BNL, FNAL –KNLs: 120M KNL core hrs @ JLAB (360M Jpsi)

  • New resources @ BNL: 300M Jpsi core-hrs : GPU or KNL

Total available: 1338M (last year: 1260M) Jpsi core-hrs

Modest increase

Available resources

slide-5
SLIDE 5

Proposals 2017-2018

The SPC has received

  • 35 Type-A proposals (27 last year)


 
 


  • 4 Type-B proposals (can be requested any Fme) (5 last year) 


These are within fluctuations are the same numbers as last year

  • 4 for Incite 0-priority Fme
  • 1 for MILC configuraFon storage
  • Rest for physics projects

slide-6
SLIDE 6

Proposals 2017-2018

  • Many of the proposals are conFnuaFon of past years -

The SPC has asked for report on progress, improvements, 
 Fmeline to finish if appropriate

  • There are a few all or mostly new efforts - 


We tried to invite those PIs to present here

  • The SPC had no direct input to the Incite proposal and has no role

in distribuFng those resources. The EC will solicit input about possible change in the Incite proposal process - roundtable tomorrow.

slide-7
SLIDE 7

Proposals 2017-2018

  • 35 Type-A proposals

Resource Request Available Req/Avail BNL BG/Q 23.4M 17.7M 132% Clusters(M) 645M 261M 249% GPU(K) 8236K 8396K 98%

KNL(Jpsi)

789M 360M 219%

New(Jpsi)

300M extra

Total

2153M Jpsi 1338 M 161%

  • 60% over-subscripFon is not bad, but the clusters and KNL are very over-

subscribed while GPUs are not even 100%;

  • The new hardware purchase should correct this imbalance.
slide-8
SLIDE 8

Distribution per field (requested) 2017-18

2017-18: IF: 446M EF: 143M NP: 1202M Th: 360M

200 400 600 800 1000 1200 1400 IF EF NP TH

Request per field (Jpsi equiv) 2017-18

Cluster GPU BNL(Q) KNL

IF EF NP TH Not including Incite

slide-9
SLIDE 9

0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 IF EF NP TH

Request per field (Jpsi equiv) 2016-17

Cluster GPU BNL(Q) KNL

Last year:

Total:(J-psi hrs) 2016-17: IF: 529M EF: 86M NP: 812M Th: 272M

IF EF NP TH

2017-18: IF: 446M EF: 143M NP: 1202M Th: 360M

slide-10
SLIDE 10

Per-cent request per field 2017-18

0.1 0.2 0.3 0.4 0.5 0.6

% Request per field

Total:(J-psi hrs) IF: 21% EF: 6.5 % NP: 56 % Th: 16.5 %

IF EF NP TH I m b a l a n c e

slide-11
SLIDE 11

Request vs availability

Total:(J-psi hrs) Clusters 250% GPU 98% KNL 120% * BG/Q 132%

* BNL new added


as KNL

100 200 300 400 500 600 700 800 900 Cluster-req Cluster-avail GPU-req GPU-avail KNL-req KNL-avail BGQ-rq BGQ-avail

Architecture request vs available

IF EF NP TH FNAL JLAB BNL

Cluster GPU KNL BG/Q request vs available I m b a l a n c e

slide-12
SLIDE 12

Proposals 2017-2018

SPC is a]empFng to raFonalize availability across all resources

  • ScienFfic merit of proposals

  • Which projects align with US HEP / NP programs ?
  • Could projects share resources or be combined ?
  • Can projects shib resource request? (clusters --> KNL or GPU)
  • What is the architecture of the new resource?
  • What is the desired balance between NP and HEP ?
  • We would like to discuss some of these issues during the roundtables
slide-13
SLIDE 13

Communication

Until now the SPC chair was not a member of the EC, creating occasional communication problems;

This is changing now, the chair of SPC will be a member of the EC form now on.

Communication between SPC and USQCD membership could also be improved

Secure website where proposals, reviews, responses and allocations are uploaded, minimizing individual e-mail

  • traffic. The website could provide forum for general

discussion and also questions, messages to SPC and EC.

Save your suggestions for the round table discussion