�ystein Gr�vlen wrote:
"AH(" == Amit Handa (JIRA) <[email protected]> writes:
AH(> [ http://issues.apache.org/jira/browse/DERBY-150?page=comments#action_12312720 ]
AH(> Amit Handa commented on DERBY-150:
AH(> ----------------------------------
AH(> To further work on this topic there are two things
AH(> 1. To identify a test case that logging consumes time cycles.
AH(> 2. And if logging is turned off as described here http://incubator.apache.org/derby/docs/adminguide/tadminlogfile.html
AH(> Does it bring the required results ? In that case it may not be a bug...
AH(> Can somebody or the submitter please file some sort of reproducible
here ?
We are talking about two different kinds of logging here. What you
are referring to is the derby.log file (aka trace or history file).
Derby-150 talks about being able to turn off the logging of database
operations which is used to roll back transactions and perform
recovery. Currently Derby first creates log records and then uses the
log records to execute the db operations. Hence, it is certainly
not straight-forward to get rid of the log generation. However, not
writing the log to disk should be possible achieve.
But turning off logging for recovery operations could hamper the
implementation of Derby's version of aries algo. Also, if the log is not
written to the disk and is maintianed elsewhere ( main memory ? ) what
do we do in case of hardware/CPU crash how does recovery happen ? We'll
have to have an architecture of failover, is that a good idea for Derby ?
~Shreyas
