A distributed cache accepts only serializable keys and values. If a service attempts to put a non-serializable object into the cache at run time, the service will receive an exception.

A distributed cache uses the disk store on the Terracotta Server Array, not on the Integration Server. You cannot configure the portion of distributed cache that resides on the Integration Server to overflow or persist to disk. When you create a distributed cache using Integration Server Administrator, the Overflow to Disk or Disk Persistent settings are disabled.

An Integration Server references a cache on the Terracotta Server Array using a fully qualified cache name. The fully qualified name of a cache consists of the name of the cache manager and the name of the cache. For example, the fully qualified name of a cache called “OrderDetails” in a cache manager called “Orders” is “Orders.OrderDetails.” If you have multiple Integration Servers that share a distributed cache, be sure that they all use the same fully qualified name for the cache.

To create a distributed cache that will be shared by multiple Integration Servers, you first create and enable the cache on one of the Integration Servers. This step registers the distributed cache on the Integration Server and also creates the cache on the Terracotta Server Array. Then you add the distributed cache to each additional Integration Server that will use the cache. When you enable the distributed cache on these Integration Servers, they will see that the cache already exists on the Terracotta Server Array and will begin using it.

Certain parameters for a distributed cache are cache-wide, meaning that when you change the parameter on one Integration Server, it affects all Integration Servers that use the cache. For information about which parameters are cache-wide, see Cache-Wide and Client-Specific Parameters for a Distributed Cache.

For a Terracotta Server Array you can set failover behavior to consistency instead of availability (the default). For information, see the section on failover tuning for guaranteed consistency in the BigMemory Max Administrator Guide.