Usenet.com

www.Usenet.com

Group Index

Comp Thread Archive from Usenet.com

<-- __Chronological__ --> <-- __Thread__ -->

Re: Berkeley-DB too fast



Michael Ubell <[EMAIL PROTECTED]> wrote:
) [EMAIL PROTECTED] wrote:
) 
)> Is there a way to force writing data to disk instead of writing it just to
)> some cache?
)>
) 
) Matthias,
) 
) So long as you specified DB_INIT_LOG and did not set the DB_TXN_NOSYNC or
) DB_TXN_WRITE_NOSYNC flags transactions are committed to disk before
) txn->commit() returns.  If your measurement was of the average commit
) time of many concurrent transactions then you are probably seeing the
) effect of our group commit logic which commits several transactions
) in 1 write to the log.  Note also that if a transaction does not perform
) any updates then txn->commit() will not write any log records.
) 

I did write some information but nonetheless the last one was the reason,
I accidently did not pass the transaction handle to DB::put.

Now an insert within a transaction takes about 50msec. That seems to be 
reasonable. I turned off harddisk cache, since information stored within
that would be lost in case of a power failure. With activated cache it
takes about 1.2msec

Matthias Meixner

-- 
Matthias Meixner                       [EMAIL PROTECTED]
Technische Universität Darmstadt
Datenbanken und Verteilte Systeme              Telefon (+49) 6151 16 6232
Wilhelminenstraße 7, D-64283 Darmstadt, Germany    Fax (+49) 6151 16 6229



<-- __Chronological__ --> <-- __Thread__ -->


Usenet.com



Please check out one of the premium Usenet Newsgroup Service Providers below for access to Usenet.