Loading...

Change Owns to Parent Ofs

Set start and due date...

Xporter

XML

Word

Printable

Details

Type: Feature
Priority: Could have
Fix Version/s: PI23
Component/s: SRCnet Workflow Verificatons
Labels:

ARTs:

SRCnet
Benefit hypothesis:

Hide

Validating the efficacy of the UKSRC profiling tool on SRCNet workloads, provides a demonstration of how all SRCs can easily attain information on computing usage for their workloads. For workload developers, this can assist in identifying where optimisations can be found within a workload.

Collecting the profiling data and analysis of the multiple workloads then also informs hardware procurement, as identifying the most common bottlenecks across workloads highlights what hardware would be most beneficial to purchase. Additionally, the knowledge of the hardware usage of various workloads informs users on which systems could run the desired workloads without requesting excessive amounts of hardware the workload can't use effectively.

The outcome from this feature will inform the compilation of Infrastructure Comput tests for V0.1

Show
Validating the efficacy of the UKSRC profiling tool on SRCNet workloads, provides a demonstration of how all SRCs can easily attain information on computing usage for their workloads. For workload developers, this can assist in identifying where optimisations can be found within a workload. Collecting the profiling data and analysis of the multiple workloads then also informs hardware procurement, as identifying the most common bottlenecks across workloads highlights what hardware would be most beneficial to purchase. Additionally, the knowledge of the hardware usage of various workloads informs users on which systems could run the desired workloads without requesting excessive amounts of hardware the workload can't use effectively. The outcome from this feature will inform the compilation of Infrastructure Comput tests for V0.1
Acceptance criteria:

Hide

1. Demonstrate UKSRC profiling tool to measure an SRCNet repository workload.

2. Measure STARScore of UKSRC / SPSRC resources (Where possible)

Show
1. Demonstrate UKSRC profiling tool to measure an SRCNet repository workload. 2. Measure STARScore of UKSRC / SPSRC resources (Where possible)
Expect Dependencies:

Intra Program
Feature Points:
2
Initial Size:
2
WSJF:
0
Epic Link:
Science Enabling v0.1 - Roadmap
Agile Teams:

Team_CORAL
Due Sprint:
Sprint 5
Story Point Burn-up:
Overdue:
Outcomes:
Hide

AC1. Demonstrate UKSRC profiling tool to measure an SRCNet repository workload.

Demo: https://confluence.skatelescope.org/pages/viewpage.action?pageId=280401971

Repo: https://gitlab.com/ska-telescope/src/src-workloads/-/tree/reframe?ref_type=heads

AC2: Measure STARScore of UKSRC / SPSRC resources (Where possible)

The way to run the workflow to apply de STARSscore at UKSRC:
https://confluence.skatelescope.org/display/SRCSC/COR-+632++%5Bprofiling%5D+Run+the+workflows+to+apply+the+STASRSscore+on+the+SP+Resources+-+part+1

Prepare computing resources required at the espSRC for applying STARS: https://confluence.skatelescope.org/display/SRCSC/COR-+631+%5Bprofiling%5D+Prepare+the+computing+resources+required+at+the+espSRC+for+the+applying+STARS

STASRS score on the SP Resources: https://confluence.skatelescope.org/display/SRCSC/COR-+632++%5Bprofiling%5D+Run+the+workflows+to+apply+the+STASRSscore+on+the+SP+Resources+-+part+1
Show
AC1. Demonstrate UKSRC profiling tool to measure an SRCNet repository workload. Demo: https://confluence.skatelescope.org/pages/viewpage.action?pageId=280401971 Repo: https://gitlab.com/ska-telescope/src/src-workloads/-/tree/reframe?ref_type=heads AC2: Measure STARScore of UKSRC / SPSRC resources (Where possible) The way to run the workflow to apply de STARSscore at UKSRC: https://confluence.skatelescope.org/display/SRCSC/COR-+632++%5Bprofiling%5D+Run+the+workflows+to+apply+the+STASRSscore+on+the+SP+Resources+-+part+1 Prepare computing resources required at the espSRC for applying STARS: https://confluence.skatelescope.org/display/SRCSC/COR-+631+%5Bprofiling%5D+Prepare+the+computing+resources+required+at+the+espSRC+for+the+applying+STARS STASRS score on the SP Resources: https://confluence.skatelescope.org/display/SRCSC/COR-+632++%5Bprofiling%5D+Run+the+workflows+to+apply+the+STASRSscore+on+the+SP+Resources+-+part+1
Resolved PI.Sprint:
24.3

Feature Checklist:

Stories Completed, Outcomes Reviewed, Demonstrated

Requirement Status:

PI24 - UNCOVERED
Labels_MIRO:
SRCNet0.1 compute-tests tests-compilation

Description

In order to measure hardware usage of the workflows used by demonstrator cases in the UKSRC, the PyProfQueue package was created. This package takes in bash scripts which it uses to submit jobs to HPC queuing systems with the needed declarations and initialisation to track hardware usage with Prometheus and Node Exporter, as well as a performance measure using Likwid. The hardware usage can be used to understand the resources that are required by a workflow, and track the Hardware usage through out the lifetime of the workload execution.

To test the robustness of this package, and adapt it to the wider communities needs, we aim to use this package with SRCNet repository workloads. By doing this, we can validate if it performs as expected on a wider range of workloads, and change it where it falls short, but we can also learn more about the metrics that other workload developers may require or desire. By combining this with a STARScore measure of UKSRC resources, we can centralise the information of SRC owned or sponsored hardware that is available in the UK, with the aim of simultaneously providing feedback for the development of the STARS suite.

Attachments

Issue Links

Child Of

SP-4870 Science Enabling v0.1 - Roadmap

Implementing

relates to

SP-4286 Add additional peak performance benchmark metrics alongside STARS

Program Backlog

mentioned in: Page Loading...; Page Loading...; Page Loading...; Page Loading...; Page Loading...; Page Loading...; Page Loading...

(4 mentioned in)

Structure

Activity

People

Assignee:: Perry, Robert

Reporter:: Keil, Marcus

Votes:: 0 Vote for this issue

Watchers:: 0 Start watching this issue

Feature Progress

Story Point Burn-up: (100.00%)

Feature Estimate: 2.0

	Issues	Story Points
To Do	0	0.0
In Progress	0	0.0
Complete	8	21.0
Total	8	21.0

Dates

Created:: 07/May/24 10:53 AM

Updated:: 1 week ago 6:43 AM

Resolved:: 15/Oct/24 6:08 AM

Due Sprint Date:: 20/Aug/24

Adapt UKSRC profiling tool for quantitative and analytical feedback