This does not seem a very smart thing to do for any standard application especially without a well tested LInK library. When it was necessary and was easy, I h

This does not seem a very smart thing to do for any standard application especially without a well tested LInK library. When it was necessary and was easy, I had local in memory caches for some specific hot data acting as another level of caching. In the simplest case it us just one variable + timestamp. Otherwise usually a Map or a simple library will do it. But substituting your caching layer with in process caching, it is usually very risky. In cases where the system cannot work without a hot cache, in memory caching can easily lead to extened downtimes of the whole system. First of all, the deployment process needs to be reengineered so the new version of the app can load the cache of the current running process. Then if there is a bug that crashes the process, you end up losing all the cache (unless like Redis you also save it on disk). Until there is a well tested library that can do that or the system really need a system like this, I am going to keep using Redis and use a second in memory caching where it is easy and low risk to do.