Re: Top 100 Maintainers List, lir-wg@localhost, routing-wg@localhost

[Joao Luis Silva Damas joao@localhost]

Hi,

depends on what you consider wrong.

Response time is definitely wrong. The system is, however, working.

The system has on some occcasions, for the last two weeks, received
messages at a rate that is faster than the system can handle.

This is partly due to the fact that a number of huge messages (containing
in the order of 10^4 objects) have been received and they request updates
of objects of which there are a big number in the database. The current
software hits severe scalability problems when this happens.

We have during the last days, devoted a lot of attention to solving these
problems by specifically hacking the software to optimize this type of
updates. We have tuned the system parameters to the limit. We are currently
in the process of trying out an update to the disk system (which to add to
all of the above, has suffered an unusual amount of physical problems).

We have also contacted the senders of the updates to try to have the
messages split and give a chance to everyone else.

We are looking at a front end for the database that will do some kind of
batch queue handling enabling control over how resources are distributed to
users, so that no single user can hog the machine.

Also, this is impacting the efforts of the reimplementation project which
is the ultimate solution for this problem.

Hope this gives you a bit of an overview of the situation.

I apologize for the incovenience this is generating for everyone but I
assure you, we are doing our best to get this thing to work as fast as
possible.

Best regards,
Joao Damas
RIPE DB Group manager
RIPE NCC

At 17:52 +0200 6/9/99, Hank Nussbacher wrote:
>On Mon, 6 Sep 1999, RIPE Database Administration wrote:
>
>Is something wrong with auto-dbm@localhost  I have not received any reply
>from that id for the past 48 hours.
>
>-Hank