STORY: As a developer bringing up my own IMAGE server, I want to minimize the resource requirements necessary to run the complete IMAGE server stack, so that I can serve as many clients as possible with the lowest possible cost.

[jeffbl]

• Profile existing server stack to find bottlenecks when serving multiple clients simultaneously
• Create scripts/tools for load testing server on an ongoing basis
• Review environments used by ML preprocessors, optimize where necessary
• Design system for caching requests, e.g., if multiple clients ask for same graphic in short time window
• Document process and best practices for running an IMAGE server (This might be significant enough to fall into another sprint?)

The goal of this story is to think broadly about where likely bottlenecks will occur, and where we can find the biggest bang for the buck in terms of making it easy and low-cost to run an IMAGE server. We are used to running on pegasus, but that is a fairly expensive machine. By optimizing, the hope is that a viable server could be run on lower-end hardware, or alternatively, scale much higher on hardware like pegasus.

[jeffbl]

Note some additional information in older issues, which can probably be closed/folded into this story, but might contain useful information on prior efforts in this direction:
#251
#266
#93
Shared-Reality-Lab/IMAGE-test-graphics#18

[jaydeepsingh25]

Server cache implementation merged and deployed on unicorn

[jaydeepsingh25]

Closing this issue as multiple issues have been created to track different items