Grenoble:Wattmetre: Difference between revisions

From Grid5000
Jump to navigation Jump to search
No edit summary
No edit summary
 
(12 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{Portal|User}}
{{Portal|User}}
; The electrical power consumption of all nodes in Grenoble is monitored by wattmetre devices from [http://www.omegawatt.fr/ Omegawatt] (same as nodes in Lyon).
; The electrical power consumption of the yeti, troll, and servan clusters in Grenoble is measured by wattmetre devices from [http://www.omegawatt.fr/ Omegawatt] (same as nodes in [[Lyon:Wattmetre|Lyon]] or [[Nancy:Wattmetre|Nancy]]).


Up to 50 measurements per second are recorded for all power supply units of all nodes.
* 50 measurements per second are recorded for every power supply unit of those clusters' nodes.
* Each troll node has a single power supply unit (PSU)
* Each yeti node and each servan node have redundant power supply units.
* yeti and troll are monitored by the device named wattmetre-1, servan is measured by the device named wattmetre-2.


For now, '''only the yeti nodes are supported by the full featured tools''' described in the [[Energy consumption monitoring tutorial]].
{{Note|text=Until 2022-09-01, the power consumption of the dahu cluster was also measured, although not available in kwollect (because measured by groups of 4 nodes). Dahu nodes aren't measured anymore due to hardware issues with the wattmetres (see the former version of this page for more information on the dahu measures).}}
 
== Dahu row measures ==
The dahu nodes are more complex in the fact that they are grouped in chassis of 4 nodes, with 2 power supply units per chassis.
 
The power consumption information for the '''dahu is for now limited to the CSV files''', as described in [[Energy_consumption_monitoring_tutorial#On_Lyon_site_and_Grenoble.27s_yeti_cluster|this section of the tutorial]].
 
The row files of measures are located here: http://wattmetre.grenoble.grid5000.fr/data/ (only accessible from within the Grid'5000 network, or using Grid'5000 VPN).
 
For instance, one CSV line of measures is:
c0:00:fc:00:01:3b:00:40:e6:02:3b:09:66:08:c6:08:35:08:03:08:50:08:13:09:ab:09:ff:ff:00:00:00:00:01:00:8a:16:82:16:a5:11:b8:11:1b:17:e1:16:35:13:15:00:f7:16:c4:16:66:13:78:13:1c:19:fc:18:0e:1c:12:1c:00:00:00:00:4a:74:c1:,
00:fc:00:01:3b:00:40:e6:02:3b:09:66:08:c6:08:35:08:03:08:50:08:13:09:ab:09:ff:ff:00:00:00:00:01:00:8a:16:82:16:a5:11:b8:11:1b:17:e1:16:35:13:15:00:f7:16:c4:16:66:13:78:13:1c:19:fc:18:0e:1c:12:1c:00:00:00:00:4a:74:,
1544796940.084479088,
OK,
230.8,210.0,219.3,205.2,200.3,207.8,226.9,241.7,-0.1,0.0,0.0,0.1,563.5,562.7,441.1,443.0,577.6,572.0,480.2,2.1,574.1,569.1,485.0,486.7,627.7,624.6,701.4,701.8,0.0,0.0
 
[[File:Dahu-Yeti.jpg|400px|right]]
''"OK"'' means that the measure is good, and is preceded by the time-stamp of the measure.
Watt values begin after the ''OK'', with 30 values as follows:
{|border=1
|-
|colspan=2| Wattmetre1-log
|-
| 1 || yeti-1 psu 1
|-
| 2 || yeti-1 psu 2
|-
| 3 || yeti-2 psu 1
|-
| 4 || yeti-2 psu 2
|-
| 5 || yeti-3 psu 1
|-
| 6 || yeti-3 psu 2
|-
| 7 || yeti-4 psu 1
|-
| 8 || yeti-4 psu 2
|-
| 9 || 0 / unused
|-
| 10 || 0 / unused
|-
| 11 || 0 / unused
|-
| 12 || 0 / unused
|-
| 13 || dahu-[1-4] psu 1
|-
| 14 || dahu-[1-4] psu 2
|-
| 15 || dahu-[5-8] psu 1
|-
| 16 || dahu-[5-8] psu 2
|-
| 17 || dahu-[9-12] psu 1
|-
| 18 || dahu-[9-12] psu 2
|-
| 19 || dahu-[13-16] psu 1
|-
| 20 || dahu-[13-16] psu 2
|-
| 21 || dahu-[17-20] psu 1
|-
| 22 || dahu-[17-20] psu 2
|-
| 23 || dahu-[21-24] psu 1
|-
| 24 || dahu-[21-24] psu 2
|-
| 25 || dahu-[25-28] psu 1
|-
| 26 || dahu-[25-28] psu 2
|-
| 27 || dahu-[29-32] psu 1
|-
| 28 || dahu-[29-32] psu 2
|-
| 29 || 0 / unused
|-
| 30 || 0 / unused
|}
 
== Additional dahu nodes, in CIMENT ==
;Additional dahu nodes (same hardware) are installed in the CIMENT (Gricad) infrastructure, but monitored with the same wattmetre system.
The CSV output of wattmetre2-log gives (after the OK word) the measures in the following order:
{|border=1
|-
|colspan=2| Wattmetre2-log
|-
| 1 || dahu[33-36] psu 1
|-
| 2 || dahu[33-36] psu 2
|-
| 3 || dahu[37-40] psu 1
|-
| 4 || dahu[37-40] psu 2
|-
| 5 || dahu[41-44] psu 1
|-
| 6 || dahu[41-44] psu 2
|-
| 7 || dahu[45-48] psu 1
|-
| 8 || dahu[45-48] psu 2
|-
| 9 || dahu[49-52] psu 1
|-
| 10 || dahu[49-52] psu 2
|-
| 11 || dahu[53-56] psu 1
|-
| 12 || dahu[53-56] psu 2
|-
| 13 || dahu[57-60] psu 1
|-
| 14 || dahu[57-60] psu 2
|-
| 15 || dahu[61-64] psu 1
|-
| 16 || dahu[61-64] psu 2
|-
| 17 || dahu[65-68] psu 1
|-
| 18 || dahu[65-68] psu 2
|-
| 19 || dahu[69-72] psu 1
|-
| 20 || dahu-[69-72] psu 2
|-
| 21 || dahu-[73-76] psu 1
|-
| 22 || dahu-[73-76] psu 2
|-
| 23 || dahu-[76-80] psu 1
|-
| 24 || dahu-[76-80] psu 2
|-
| 25 || bigfoot1 psu 1
|-
| 26 || bigfoot1 psu 2
|-
| 27 || bigfoot2 psu 1
|-
| 28 || bigfoot2 psu 2
|-
| 29 || bigfoot3 psu 1
|-
| 30 || bigfoot3 psu 2
|}
* The CIMENT nodes have to be reserved and managed in the CIMENT (Gricad) infrastructure, with the CIMENT tools (https://gricad.univ-grenoble-alpes.fr/thematiques/calcul-intensif-ciment), but measures are to be retrieved in Grid'5000 (like for Grid'5000 nodes).
* For experiments involving both the Grid'5000 machines and the CIMENT machines, please contact the staff (support-staff@lists.grid5000.fr).
* The 3 Bigfoot nodes are GPU nodes.

Latest revision as of 16:02, 7 September 2022

The electrical power consumption of the yeti, troll, and servan clusters in Grenoble is measured by wattmetre devices from Omegawatt (same as nodes in Lyon or Nancy).
  • 50 measurements per second are recorded for every power supply unit of those clusters' nodes.
  • Each troll node has a single power supply unit (PSU)
  • Each yeti node and each servan node have redundant power supply units.
  • yeti and troll are monitored by the device named wattmetre-1, servan is measured by the device named wattmetre-2.
Note.png Note

Until 2022-09-01, the power consumption of the dahu cluster was also measured, although not available in kwollect (because measured by groups of 4 nodes). Dahu nodes aren't measured anymore due to hardware issues with the wattmetres (see the former version of this page for more information on the dahu measures).