Is your feature request related to a problem? Please describe.
In order to achieve better serving performance following changes has to be done:

1. Preload all spec during serving start-up process into cache.
2. Refresh the cache asynchronously

Additional context
Some client service has a very strict deadline which value is less than what it takes to fetch a spec from core. Currently, spec is stored in cache but populating it is still done synchronously during processing of incoming request. This introduce additional latency to feature request when the corresponding spec is not available in cache.