[ZODB-Dev] filestorage open latency

[Barry A. Warsaw]

>>>>> "DM" == David McCusker <david@treedragon.com> writes:

    DM> Is open latency for Berkeley db stores less?

In general, it probably is, since all the metadata is readily
available in Berkeley btrees so there's no seek-n-search necessary.
Plus, only transactional undo is supported in the Berkeley Full
storage.

The open latency is completely dependent on Berkeley's inherent
latency in opening about 20 tables, including whether any recovery
must be done.  Assuming everything's nice and checkpointed, it should
be pretty darn quick, but I haven't measured it since the rewrite.

-Barry