search

uzh / vm-mad

Dynamically grow or shrink GridEngine clusters using cloud-based nodes
https://arxiv.org/abs/1302.2529
Apache License 2.0
3 stars 2 forks source link

Modify `simul.py` so that it reads data from a file in the CSV "useful data" format #4

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
Modify the `simul.py` class so that it reads the workload from
the `.csv` file created by the `distil.py` script (see Issue 3),
instead of generating a random one.  

This also entails modifying it so that jobs are changed from PENDING
to RUNNING status only after their submit time.

Original issue reported on code.google.com by riccardo.murri@gmail.com on 30 Jan 2012 at 8:36

GoogleCodeExporter commented 9 years ago

Original comment by riccardo.murri@gmail.com on 30 Jan 2012 at 8:41

GoogleCodeExporter commented 9 years ago 
Hi R & C,

I've already completed this issue. Some fine tuning is now needed, in 
particularly: 
- I use the accounting data form our SGE cluster which first started on  "25 
Aug 2010 10:54:54". 
- In the CSV file created by distil.py the format is like this: 
  jobid, submit_time, duration
Thus, we have the history for all the jobs from the beginning till nowadays. 

Questions:
How should a realistic simulation look like? Better:
 - What should be the time period (from the CSV) taken on each iteration 
of the orchestrator: 1 hour, 12h hours, 1d...1m, etc?   
 - What about the scheduling interval of the orchestrator?
 - maxVms?
 - delay? 

We'll have some fun on Tuesday :)

Tyanko

Original comment by tyanko.a...@gmail.com on 3 Feb 2012 at 3:45

GoogleCodeExporter commented 9 years ago 
| Comment #2 on issue 4 by tyanko.alexiev:
| How should a realistic simulation look like? Better:
|  - What should be the time period (from the CSV) taken on each iteration
| of the orchestrator: 1 hour, 12h hours, 1d...1m, etc?
|  - What about the scheduling interval of the orchestrator?
|  - maxVms?
|  - delay?

These should all be configurable settings.  One of the purposes of the
simulator is to let us experiment and optimize the settings given a
real workload trace.

Original comment by riccardo.murri@gmail.com on 3 Feb 2012 at 3:52

GoogleCodeExporter commented 9 years ago 
Now all these can be passed to the simulator as configurable variables.

Original comment by tyanko.a...@gmail.com on 7 Feb 2012 at 9:07