Reduced Abaqus Token Usage within Automated Isight Studies
Starting with Abaqus 6.10 and Isight 4.5, SIMULIA introduced a new token usage policy to encourage use of Abaqus and Isight together for design exploration studies. The new system reduces analysis token usage per run when multiple Abaqus runs are executed within an Isight study. This approach makes it is more cost effective to run variations on Abaqus models using Isight than if design alternatives are explored manually. Token reductions apply for as few as 2 Abaqus runs executed within an Isight study and reach a threshold 60% target reduction when 10 to 32 runs are executed depending on the number of runs executed concurrently.
To take advantage of the token reduction, customers must use Isight 4.5 (or later) and Abaqus 6.10 (or later) and use the Abaqus component for Isight. Running Abaqus through custom components or scripts does not currently support token reductions. Further, token reductions apply only for automatically generated design points, meaning, the variation in parameter values for each Abaqus run must be generated automatically by an Isight process component, and may not be modified outside the process component.
When running a study with Isight, the Abaqus token usage per run decreases with the number of runs in the study, to a threshold representing a 60% target reduction from the default token usage. Token usage is reduced faster when Abaqus runs are submitted to execute concurrently by Isight. Two main factors control the token usage; 1) the cumulative number of Abaqus runs in the study and 2) the number of concurrent Abaqus runs submitted by Isight. Isight submits Abaqus runs in groups, referred to as “batches” to execute concurrently. Token reductions for batches submitted sequentially over the course of a study, start at 0% for the first batch and increase with successive batches. As the batch size and batch number increase, the Abaqus token usage decreases to a threshold value representing a target equal to 60% of the baseline Abaqus token usage.
The graph below qualitatively indicates the reduction of Abaqus token usage as a percentage of the baseline token usage for a comparable Abaqus run outside of Isight. These curves have been smoothed to remove the effect of granularity associated with integer token values. This approach benefits small accounts by freeing up tokens for other work while longer studies are executing. Large accounts with resources to execute more parallel runs can turn around runs faster using fewer Abaqus tokens.
To learn more, please read the attached PDF file.
