Caching functionality for a function in TIP (SBO)

[Konohana0608]

User Story
As a TIP user, I want to utilize a shared cache for function results to significantly reduce execution time. The cache should be accessible across all users and all deployments within our TIP user base.

Objective
The goal is to implement a shared caching mechanism for a computationally expensive function.

• Each computed dataset sample (i.e., an output generated by the function) is approximately 500 KB in size.
• The system will handle several thousand computed samples.
• Samples are uniquely indexed using model name, target tissue, and other key parameters.
• When a user computes a dataset sample for the first time, it should be stored in the shared cache, allowing future users to retrieve it instantly instead of recomputing it.

Expected Benefit
The most time-consuming process in our surrogate-based optimization workflow is evaluating training dataset samples to build the model. By caching these computed samples:

• The first user to compute a specific sample will store it in the cache.
• Subsequent users needing the same sample can retrieve it instantly, avoiding redundant computations.
  This will lead to significant time savings across all users.

Technical Considerations

• The cache should be shared globally across all users and deployments.
• The caching mechanism should ensure data integrity and uniqueness based on the defined indexing criteria.
• The system should handle thousands of cached computed samples (relatively) efficiently.

Would love to hear thoughts on potential caching solutions and best practices for implementation!