I know that we can put files as caches entries in infinispan (as the Lucene Directory of Infinispan). But handling big binaries is another kind of animal: I cannot have a 4gb file in memory.

So I was thinking if there is a way to stream instead of read a cache entry ... is this possible?  I was thinking to have a cache for big binaries with no memory space, so everything will be keep in the secondary storage (database or file system). Then every time that someone get a key from this cache it can obtain a 'DataHandler' or 'InputStream' to read the big binary.

Why?

Our software use the cluster to read almost all the business data but files attached to business objects. The idea is to have all the files that our software use on the IMDG instead of having another software doing this for us (i.e. distributed file system or a network file system as our current implementation).

Thanks in advance,