AWS ElastiCache configuration endpoint becomes unreachable after some time

We have a MVC ASP.NET App and we get a really high number of requests per second, like typically between 1500 to 2000 and at peak times during week we may get 4000 requests in one second according to the IIS Logs. We were trying to use AWS Elasticache to keep certain basic data about the client that is communicating with the WebServer. It is not a browser but a Windows Service that calls into some WebAPIs. The security groups and the VPC settings seems to be all right as in the start I can telnet to each nodes including the configuration endpoint.

After successfully adding information to the cache or after successfully retrieving information from the cache after few minutes of high number of concurrent requests the configuration End point becomes unavailable. I can not telnet to the configuration end point on port 11211. Has anyone else also faced such problems?

Isn't the whole point of caching supposed to be handling really high traffic without having to go to the Database every time and thus reducing the latency? We are indeed using the AutoDiscovery of nodes and we are using the "ElastiCache Cluster Client" from Amazon. And we have tried upto 4 nodes, tried with various types of node from micro to all the way upto xlarge. But it does become unresponsive after a while. Are we supposed to throttle the number of requests going to cluster?

Any help / pointers will be really appreciated.