Joint WLCG & HSF Workshop 2018

Session

Performance and Cost Modeling Introduction

28 Mar 2018, 11:00

Centro Congessi Federico II (Aula Magna; Aula A); Hotel Vesuvio (Napoli, Italy)

Conveners

Performance and Cost Modeling Introduction

• Jose Flix Molina (Centro de Investigaciones Energéti cas Medioambientales y Tecno)
• Andrea Sciaba (CERN)
• Markus Schulz (CERN)

Performance and Cost Modeling Introduction: Technical Session

• Jose Flix Molina (Centro de Investigaciones Energéti cas Medioambientales y Tecno)
• Markus Schulz (CERN)
• Andrea Sciaba (CERN)

Description

Overview of the Working Group, Mandate, Activities and Status
Mainly for non working group members

70. Overview of the Working Group, Mandate, Activities and Status

Jose Flix Molina (Centro de Investigaciones Energéti cas Medioambientales y Tecno)

28/03/2018, 11:00

Mainly for non working group members. Intended to raise interest and stimulate contributions.

72. Most important workloads and how to run them

Johannes Elmsheuser (Brookhaven National Laboratory (US))

28/03/2018, 11:30

Reason why these workloads are “important” in matters of cost.
Focus on Run2 experience
Run3 and HL-LHC will be covered in the afternoon session
Summarise what we have, list of workloads
Overview of the experience gained running these workloads (especially from non experiment members view)

73. Metrics and Measurements

Gareth Douglas Roy (University of Glasgow (GB))

28/03/2018, 11:45

The metrics relevant for the workloads.
Status of the measurement of these quantities for the different workloads.
More detailed discussions either in the afternoon session or during dedicated meetings

74. Cost evaluation process

Catherine Biscarat (LPSC Grenoble, IN2P3/CNRS)

28/03/2018, 12:05

Mapping the cost model to site relevant costs
Walk through an simple example, motivation etc.

Afternoon: Concrete example calculations, where to go from here, discussions

78. Tools and Techniques for Performance/Cost Measurements

Andrea Valassi (CERN)

28/03/2018, 13:30

A presentation of tools and techniques needed to measure the performance metrics.
15' presentation followed by 15' discussion

79. Evolution of the critical workloads in Run3 and HL-LHC

David Lange (Princeton University (US))

28/03/2018, 14:00

Run3 and HL-LHC , What will change?
How can we “fake” in a more realistic way these conditions until we have the code that handles data for these periods?
Will there be an impact of the change of global architecture and what will it be, how can we quantify it?
Starting with a 3-4 slide introduction, then interactive work and discussion.

80. Cost evaluation process, example calculation

Renaud Vernet (Centre National de la Recherche Scientifique (FR))

28/03/2018, 14:30

Walk through the example
Discussion on where to go from here

75. Common resource calculation model: Brainstorming/Hackathon

Andrea Sartirana (Centre National de la Recherche Scientifique (FR))

28/03/2018, 14:50

How to estimate resource needs for run periods.
From several Excel sheets to a generic Python code with different configuration parameters by experiment.

Starting from a simplistic Python script (needs to be prepared)
Discussion and work on the code

82. Summary, Follow up

Markus Schulz (CERN)

28/03/2018, 15:20

First attempt at giving a high-level summary of the technical session

71. Importance of a common vocabulary

Andrea Sciaba or Daniele tbc

Short introduction of how we call things ( workloads, workflows etc.)
Table of workloads (steps and chains) and the transformations that they do