By Mayuresh A Nirhali-Oracle on Dec 21, 2006
This topic on FTPOnline talks
about current challenges in SOA development in the context of Databases. The article has described the need for Highly Available Distributed database really very well. The cluster solutions today are very sophisticated, but they are not
targeted towards databases. This disconnect brings up a lot of scalability and performance issues which could be solved with equally sophisticated database implementations which are targetted for such
clustered techonologies today.
Sun has been working on this integration for quite some time now and there exists a solution which could be very useful for the current SOA database bottlenecks. The solution is HADB, the old clustra (for the folks who know the clustra DB). HADB maintains independent copies of the fragmented data on server nodes and all this is completely transparent to the Application/user. Also, HADB provides 99.999 % (the 5 9's) availbility with a clustering distributed database solution. Due to the clustering nature, there are some hardware and network constraints to this solution, but then it is targetted only for 5 9's availability and hence, it is not for small-scale business apps.
Currently Sun Java Application Server provides HADB for session failover. Mani talks about how to use HADB in his blog. The famous Honeycomb project also uses HADB internally to satisfy their availability needs.