Distributed data visited on a Storm cluster and how does MemoryMapState store distributed data?

I am trying to design a topology that can have some so called "globals" just like the Distributed caching in storm question . I know it's better to use some sort of distributed cache store like Memcache or Redis, but I still have a few questions:

• What's the best way to use data caching? It looks like I shouldn't use caching directly in the function execute
  
  or nextTuple
  
  should I spawn a thread in the spout / bolt to periodically receive this data?

• MemoryMapState
  
  is a "state" used to store a permanent result in a tedent storm. He said he MemoryMapState
  
  uses ConcurrentHashMap
  
  caching for storing data as the codes in class MemoryMapState
  
  show below

  static ConcurrentHashMap<String, Map<List<Object>, Object>> _dbs = 
              new ConcurrentHashMap<String, Map<List<Object>, Object>>();
  
        
  
          
          
          
        
  
      

  So how can you ConcurrentHashMap
  
  store and query data among different machines in a Storm cluster?