XMLWordPrintable

    Details

    • Type: New Feature
    • Status: Accepted
    • Resolution: Unresolved
    • Affects Version/s: BIGDATA_RELEASE_1_2_2
    • Fix Version/s: None
    • Component/s: HAJournalServer
    • Labels:
      None

      Description

      It should be possible to create a highly available replication cluster using the HAJournalServer and the MemoryManager. HA would provide write ahead logs (HALogs), snapshots, and replication onto a quorum of nodes. The nodes themselves would use the native heap to store the data (plus write ahead log files). Restart would require reading the most recent snapshot into memory and then replaying the HA log files. The same mechanisms would exist for incremental resynchronization after a transient disconnect.

      This would provide a very high performance replication cluster, especially once combined with support for concurrent unisolated transactions against the Journal (see the relevant ticket). Query performance would especially benefit since all data would be in RAM. This deployment model would target high-mem clusters (server grade machines with 128G to 1TB of RAM) and many cores (24+ cores).

        Attachments

          Activity

            People

            Assignee:
            martyncutcher martyncutcher
            Reporter:
            bryanthompson bryanthompson
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

              Dates

              Created:
              Updated: