husky-parul
commented
8 months ago @jiere
As it shown, does it mean the "Y-axis" value of the black curve is "power" and the "Y-axis" value of the red curve is "Total power reduction"?
In the graph The black curve is plotted against the values on the left hand side of the y axis. It represents the power values of the node. The red curve represents power reduction when the cluster was using packing strategy, the same principle PEAKS propose. It is plotted against the values on the right hand side Y axis.
How could the power reduction data come out? IIUC, the data comes from the "peaks-default.png". Since the diagram title is "PEAKs vs Default", this diagram is part of the methodology, how could PEAKS scheduling happen before the born of methodology, what's the PEAKS scheduling mechanism at first?
We used the following scheme:
- the default kube scheduler spreads the load, and
- PEAKS packs the nodes.
Asser Tantawi conducted these tests. I will request for more information on the implementation of the PEAKS mechanism
Another question is about the meaning of the "power profile". What's the meaning of the Y-axis value of each point on the black curve? Idle power without any workload scheduled onto node by PEAKS or Default?
@ Asser Tantawi
It's my honor to touch this great project, some questions raised above, please kindly clarify them, thanks.