cliffmoon edited this page Sep 12, 2010 · 2 revisions

Milestones Targeted for 1.0

  • Hinted handoff
    • enables a cluster to be always writeable as long as one node is up.
  • Incremental backups
    • Allow one cluster to incrementally be backed up to another cluster
      with similar partitioning constants
  • Buffered writes
    • Enable a mode with reduced durability where writes are buffered and
      flushed to disk in the background
  • Avg Latency and 99.9% latency within the same order of magnitude as
    those described in the paper
  • Load distribution efficiency above 0.95 (mean load / max load)
  • Integration test suite for distributed failure scenarios