Continuent announces p/cluster 2006 and Sequoia 2.6


News: Continuent announces p/cluster 2006 and Sequoia 2.6

  1. Continuent has announced the general availability of p/cluster, a high availability solution for PostgreSQL. It integrates seamlessly with any Java application (or app server) as if it were using a single PostgreSQL instance. No application changes are required, just the datasource definition has to be configured with a new JDBC URL.

    The clustering features offer high availability with transparent failover and load balancing.
    p/cluster is based on the Sequoia open source technology formerly known as C-JDBC. Sequoia is a generic ANSI SQL clustering technology based on JDBC. Sequoia 2.6 has been released on New features include transparent user management, support for foreign keys, parallel recovery properly scheduling queries to respect foreign key constraints, support for transparent failover with automatic retrieval of last results on write requests, commit, and rollback. Support has also been added for automatic recovery log resync and some new experimental features include semantic support for stored procedures.

    More information on the p/cluster specific features can be found in "p/cluster 2006 is now generally available!" or in the press release on BusinessWire, "Continuent Ships Database Virtualization Solution for PostgreSQL."

    Threaded Messages (2)

  2. Write Ordering[ Go to top ]

    If I understand correctly this c-jdbc replicates databases using state-machine replication, i.e. it executes the same writes at all instances, which would mean that it has to ensure the same serialization order at each instance. Does it do this by executing one write at a time?


    Enjoy the Fastest Known Reliable Multicast Protocol with Total Ordering

    .. and the World's First Pure-Java Terminal Driver
  3. Write Ordering[ Go to top ]

    Does it do this by executing one write at a time?

    Excellent question!
    Non-conflicting writes are executed in parallel (conflict detection is based on table accesses) and conflicting writes are sent in a serializable order to the backends.
    You can also improve performance using partial replication to replicate tables on different subsets of backends to improve the distribution of queries and improve parallelism.