1 / 7

Overview of Computers in Use at WGNE Meeting, Tokyo, 9 Oct 2018

Günther Zängl provides an overview of computers in use, usage of computing resources, deterministic vs ensemble applications, data assimilation systems, and general trends at the WGNE meeting in Tokyo on 9 Oct 2018.

belmore
Download Presentation

Overview of Computers in Use at WGNE Meeting, Tokyo, 9 Oct 2018

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. WGNE Computer table WGNE meeting, Tokyo, 9 Oct. 2018Günther Zängl

  2. Overview • Overview of computers in use • Usage of computing resources: global vs. regional • Deterministic vs. ensemble applications • Data assimilation systems • General trends

  3. Overview of computers in use at WGNE members • Cray XC: ECMWF, UKMO, DWD, NCEP, NRL, CPTEC/INPE, KMA, HMC Russia (since very recently; formerly SGI, RSC) • IBM: CMC, CMA, NCMRWF (India) • Bull: MF • Dell: NRL • Hitachi: JMA • Oracle: BoM Questions for discussion • How to homogenize the numbers on sustained/peak performance? Are they still useful? Peak perf. is not representative for our model codes, measuring sustained perf. requires reliable hardware counters • Suggestion: Provide at least processor type, #nodes, #cores/node

  4. Usage of computer resources: global vs. regional Criterion: Number of cores used for deterministic forecasts: • regional >> global: MF, NCEP, CMC, JMA • regional  global: DWD, HMC, NRL, CPTEC/INPE, BoM • global >> regional: UKMO, CMA, KMA, NCMRWF • global forecasts only: ECMWF Please send updates if not correct or outdated! Remark: • Partitioning of total daily resource usage cannot be inferred from WGNE table, neither for global vs. regional nor for DET vs. EPS Question for discussion: • Shall we include such information in the table?

  5. Deterministic vs. ensemble applications Global forecasting systems • EPS mesh size <= 2* DET mesh size: ECMWF, UKMO, MF, HMC, NRL, JMA, CMA, NCMRWF • EPS mesh size > 2* DET mesh size: DWD, NCEP, CMC, CPTEC/INPE, KMA, BoM Regional forecasting systems • Same mesh size for DET and EPS: DWD, CMA • EPS mesh size <= 2* DET mesh size: UKMO, MF • EPS mesh size > 2* DET mesh size: CMC • EPS for subset of model domains: HMC, NCEP, NRL, JMA, KMA, BoM Clear trend: Importance of ensemble systems has grown in the last years

  6. Data assimilation systems Global forecasting systems • 4D-Var / Hybrid 4D-Var: ECMWF, UKMO, NCEP, NRL, KMA. NCMRWF, MF, JMA, CMA • 4D-En-Var: CMC • Hybrid 3D-Var / EnKF: DWD, BoM • 3D-Var: HMC, CPTEC/INPE Regional forecasting systems • 4D-Var: UKMO, NRL, CMC, JMA, KMA, NCMRWF, BoM • 4D-En-Var: CMC • 3D-Var: MF, CPTEC/INPE, JMA • Hybrid 3D-Var / EnKF: NCEP, CMA, BoM • LETKF: DWD

  7. General trends to be observed • Growing fraction of resources spent in ensemble systems, both at a global scale and at convection-permitting scales • This is not surprising as longer-range forecasts and convection forecasts take much benefit from ensembles (or are even meaningless without) • NWP relies not as much on “strong scaling” as climate modeling • Applications including aspects of atmospheric composition (e.g. pollen, mineral dust, volcanic ash) are increasing

More Related