Soft state is useful for efficiency because of the eventual consistency model such as short-lived user sessions, stored aggregates and transformations on large datasets and general purpose write-through caches for files and database records.
Whilst soft state is lost or made unavailable due to service
instance crashes and overloads, reconstructing it through user
interaction or third-tier re-access can be expensive in terms of time
and resources [16]. User can afford database to be consistent over
time by synchronizing information between different database
nodes. They can Cache data (soft state) and use it later to increase
the database response time. They may be having a number
of database nodes with distributed data to be highly available
(partition tolerance).
Many options exist for adding high availability of programs that
manipulate soft state and these can be broadly classified into three
categories