We have deployed Kea 1.3 in our environment with ~300 subnets using the
shared networks capability and I would like to provide some feedback on
performance. Note: We exclusively use host reservations in a MySQL
database. There are no dynamic address pools. Leases are stored in a file.
We are at point where everything is working fine, but I am concerned
that Kea will probably not scale well with an increasing amount of
subnets in its current design. When enabling the debug log, I can see
that Kea queries the database multiple times for each DHCP request.
There is one query for each configured subnet. This has quite the impact
on processing time as each request causes 300 database queries in our
scenario. Queries are small and the query cache keeps disk I/O at a
minimum, but CPU utilization is fairly high for Kea and MySQL.
If it is possible at all, it would probably be better to have Kea do
only one query for the whole hosts table instead of separate queries for
each subnet. If you any other suggestions for me, I'd appreciate that as
well.
Thanks.
--
Kind regards,
Tino Lehnig
_______________________________________________
Kea-users mailing list
[email protected]
https://lists.isc.org/mailman/listinfo/kea-users
