Cloud Scheduler: Estimate Override

Normally, when you are running jobs on the cloud, the standard workflow would be to open the cloud Scheduler and estimate the model. There can be situations when your system may not have enough resource (RAM) to perform the estimate.

If this is the case, you can override the estimate and just run the job by manually requesting a specific amound of RAM. This can only be done is the Advanced Cloud Job Submission option has been enabled beforehand in the settings.

OnScale settings can be accessed through the cog icon in the right top hand corner of the application. The Configuration option takes you to a dialogue which contains all settings for all aspects of the application.

The setting in Cloud Account Configuration must be enabled to allow override estimate feature.

Once it has been enabled simply click 'Run on Cloud' to open the cloud scheduler the option to enable the Estimate Override settings should be at the bottom of the scheduler.

Things to watch out for when using the the estimate override option:

  • Max Core Hour Allowance blocks that amount, if you have 200CH in your account and block 20 CH, then those 20 CH will be allocated to that simulation as a minimum. If the simulation costs more than 20 CH, it will use the remaining CH in your account to continue the simulation if available until it reaches the MAX CH limit (See Note* Below).
  • RAM Assignments need to be large enough or the model will fail.
  • Always try an estimate before using the Estimate Override option.

Each CPU node will have a maximum amount of RAM available for simulation:

CPUs Maximum RAM Assignment
2 15.2
4 31.2
8 63.2
16 127.2
32 255.5

*Note: New accounts have a Core Hour limit (Max CH) of 150CH which controls how much core hours can be used for a single job. If you run a simulation that exceeds 150 CH you may find it fails because this maximum limit has been reached this can be changed on the portal

The estimate override is also necessary when using any linked cloud files. Click here for more information on cloud linked files.