|
Enterprise TSpaces is
a further development of the stand-alone version of TSpaces that
provides TSpaces with enterprise required facilities such as fault-tolerance
and scalability.
Enterprise TSpaces provides
fault-tolerance on different abstraction levels. By using operation
replication between TSpaces servers in so-called replication groups
the availability of tuples increases with the number of TSpaces
servers. This increases the tolerance to failures of TSpaces servers
and thereby eliminating the single point of failures that TSpaces
currently suffers from. At a TSpaces server level each server is
able to fully recover from a crash, that is, it is able to recover
its internal state and reintegrate into the replication group. Reintegration
means that the TSpaces server is able to reestablish network connections
to the other TSpaces servers in the replication group and reestablish
consistency. The latter is achieved using a hybrid replica update
protocol.
Scalability is achieved
by partitioning the tuples in the tuplespaces between multiple partitions
-- replication groups -- each having one or more TSpaces servers.
In addition to
providing fault-tolerance and scalability Enterprise TSpaces
provides support for both of these properties to be
dynamically adjusted at runtime.
The former is archieved by dynamically increasing or decreasing
the number of TSpaces servers in a replication group. The latter
is archieved by dynamically increasing or decreasing the number
of tuplespace partitions.
NOTE: Version 1 of Enterprise TSpaces, excluding
the replication work, will be available in early 2003. We
had a group reorganization and redirection in 2002, which
stalled the replication work for the time being. We will
release the more robust Version 3 TSpaces engine in early
2003, then try to find resources to finish the replication
work. Our apologies for those of you who have been waiting
for ETS to arrive -- we were also anxious to see it be
released. Dated, 11/18/2002
|
