Ticket 3996

Summary:	Read/Write Speed in Sacct
Product:	Slurm	Reporter:	Paul Edmon <pedmon>
Component:	Profiling	Assignee:	Unassigned Developer <dev-unassigned>
Status:	OPEN ---	QA Contact:
Severity:	5 - Enhancement
Priority:	---
Version:	17.02.6
Hardware:	Linux
OS:	Linux
Site:	Harvard University	Alineos Sites:	---
Atos/Eviden Sites:	---	Confidential Site:	---
Coreweave sites:	---	Cray Sites:	---
DS9 clusters:	---	HPCnow Sites:	---
HPE Sites:	---	IBM Sites:	---
NOAA SIte:	---	NoveTech Sites:	---
Nvidia HWinf-CS Sites:	---	OCF Sites:	---
Recursion Pharma Sites:	---	SFW Sites:	---
SNIC sites:	---	Tzag Elita Sites:	---
Linux Distro:	---	Machine Name:
CLE Version:		Version Fixed:
Target Release:	---	DevPrio:	---
Emory-Cloud Sites:	---

Description Paul Edmon 2017-07-14 12:00:52 MDT

I noticed that sacct and sstat have stats for how much data was read and written but not info about average and max speed that this was done at.  This would be a very handy category to add to the database, especially to diagnose IO woes.

Comment 1 Tim Wickberg 2017-07-14 17:24:14 MDT

Remarking as an enhancement request.

It's an interesting idea.

For these calculations, would the value sampled at AcctGatherNodeFreq be sufficient? I'm not aware of a source for peak / average speed that is already available through the Linux kernel, so we'd need to derive this value internally somehow.

Comment 2 Paul Edmon 2017-07-14 19:16:06 MDT

I think that would be sufficient.  I think any more accurate measure 
would need an actual profiling tool.  This would be just for quick 
estimates.

-Paul Edmon-


On 7/14/2017 7:24 PM, bugs@schedmd.com wrote:
> Tim Wickberg <mailto:tim@schedmd.com> changed bug 3996 
> <https://bugs.schedmd.com/show_bug.cgi?id=3996>
> What 	Removed 	Added
> Severity 	4 - Minor Issue 	5 - Enhancement
> Assignee 	support@schedmd.com 	dev-unassigned@schedmd.com
>
> *Comment # 1 <https://bugs.schedmd.com/show_bug.cgi?id=3996#c1> on bug 
> 3996 <https://bugs.schedmd.com/show_bug.cgi?id=3996> from Tim Wickberg 
> <mailto:tim@schedmd.com> *
> Remarking as an enhancement request.
>
> It's an interesting idea.
>
> For these calculations, would the value sampled at AcctGatherNodeFreq be
> sufficient? I'm not aware of a source for peak / average speed that is already
> available through the Linux kernel, so we'd need to derive this value
> internally somehow.
> ------------------------------------------------------------------------
> You are receiving this mail because:
>
>   * You reported the bug.
>