Hi,

I have the results from the tests (below).  Alot of the time is spent in
checking file permissions and locking the file (40 %).


Inmem

% time     seconds  usecs/call     calls    errors syscall

------ ----------- ----------- --------- --------- ----------------

 28.53    0.124727           1    118524           write

 23.42    0.102382           0    414624           gettimeofday

 20.78    0.090840           1     76513           read

 13.72    0.059977           0    191255           _llseek



Db in tmpfs

% time     seconds  usecs/call     calls    errors syscall

------ ----------- ----------- --------- --------- ----------------

 21.83    0.257073           1    263306           write

 21.18    0.249488           1    179148    179148 access

 20.61    0.242725           0    509292           fcntl64

 13.04    0.153551           0    448720           _llseek

  9.44    0.111194           1    189370           read

  8.16    0.096124           0    414624           gettimeofday
  2.55    0.030000        3750         8           fdatasync

On Tue, Aug 9, 2011 at 7:34 PM, Pavel Ivanov <paiva...@gmail.com> wrote:

> > Journal mode is WAL
>
> I believe in-memory database can't have journal mode WAL. So you
> compare completely different settings.
>
>
> Pavel
>
>
> On Tue, Aug 9, 2011 at 5:15 AM,  <sreekumar...@gmail.com> wrote:
> >
> > Journal mode is WAL
> >
> >
> > ------Original Message------
> > From: Roger Binns
> > Sender: sqlite-users-boun...@sqlite.org
> > To: General Discussion of SQLite Database
> > ReplyTo: General Discussion of SQLite Database
> > Subject: Re: [sqlite] In memory v/s tmpfs
> > Sent: Aug 9, 2011 2:42 PM
> >
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > On 08/08/2011 06:34 PM, sreekumar...@gmail.com wrote:
> >> From the point of view of performance, I expected similar performance ,
> tmpfs being a little slower due to filesystem overhead. However, the
> operations on tmpfs was much slower than expected.
> >
> > Using tmpfs requires many kernel calls which is considerably more effort
> > than the occasional malloc call.  Additionally files have to be locked,
> > journals made etc (you didn't mention your journal setting).
> >
> > Roger
> > -----BEGIN PGP SIGNATURE-----
> > Version: GnuPG v1.4.11 (GNU/Linux)
> > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
> >
> > iEYEARECAAYFAk5A+g0ACgkQmOOfHg372QTR8ACgqNeeuOxHRy7+hMH5RY/OAyV2
> > Wq0AoMaSRtoFN4obCgmgHlpHthd9z5Zp
> > =pkJt
> > -----END PGP SIGNATURE-----
> > _______________________________________________
> > sqlite-users mailing list
> > sqlite-users@sqlite.org
> > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
> >
> >
> > Sent from BlackBerry® on Airtel
> > _______________________________________________
> > sqlite-users mailing list
> > sqlite-users@sqlite.org
> > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
> >
>
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users

Reply via email to