gnunet-developers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [GNUnet-developers] Faster (sequential ) inserting of records in MyS


From: Christian Grothoff
Subject: Re: [GNUnet-developers] Faster (sequential ) inserting of records in MySQL.
Date: Sat, 5 Apr 2003 13:20:55 -0500
User-agent: KMail/1.4.3

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Personally, I think that while good performance is nice and all, users may be 
quite confused if the file that they put in does not instantly appear.  Also, 
the implementation would get quite a bit more complicated, which is probably 
the best argument against it. After all, users are not going to insert 
content all the time; it's much more likely that they run gnunet-insert-multi 
once and are done with it (until we break the datastore format again... :-).

I would vote to focus on more pressing problems (e.g. bugs, namespaces, 
directories, http transport, IPv6, documentation) before looking after this 
type of very minor performance problems. I can index 50 MB in 90s into the 
gdbm database, if mysql is even faster, I can hardly see why we should work 
on insertion speed at this point.

Christian

On Saturday 05 April 2003 10:42 am, Igor Wronsky wrote:
> On Sat, 5 Apr 2003, Jan Marco Alkema wrote:
> > /* It turns out that for bulk inserts, for MySQL the "LOAD DATA"
> > statement is by far the fastest method. In tests, it was about 5 times
> > faster than INSERTs with multiple VALUE lists.
> > Maybe the use of the "LOAD DATA" command (+ Unixsocket) can speed up
> > gnunet?
>
> Maybe it could. There's only one slightly problematic issue with this
> that I can think of: lack of reliability in a trivial solution. On the
> other hand, if we allow for the situation where inserted blocks
> are not immediately available for requesting and accept that in the
> case of crash some fresh blocks can get lost, then its rather
> straightforward to implement the batch inserting, and goes
> similarly to what Iceman outlined there.
>
> Any implementation that'd handle the two cases would probably
> be a bit tedious to code.
>
>
> Igor
>
>
> _______________________________________________
> GNUnet-developers mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gnunet-developers
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE+jx6H9tNtMeXQLkIRApBTAKCWzYiBsFnG/tLfiY00GAfbKBRMnACeNPF4
tkTbLey2E+4VcDzAMNGZuwo=
=7F6y
-----END PGP SIGNATURE-----





reply via email to

[Prev in Thread] Current Thread [Next in Thread]