On Wed, Jul 20, 2016 at 4:23 AM, Ask Bjørn Hansen <[email protected]> wrote:
>
> Hi everyone,
>
> There are some zones that have way too few servers for way too many
clients so they basically never get to work. Someone joins the zone and in
short order they get kicked out or leave because there’s just too many
queries. When there are no servers in the zone the queries from that
country (mostly) fall back up to the continent zone.
>
> In the past we had a few zones that were in a similar situation and some
of you had your servers outside that country added to the country zone.
>
> I’m thinking of making this more automated, so when there are too few
servers in a country, add some (random-ish) servers from the rest of the
region or even the world.
>
> We could have a special flag that designates a server to be used in this
situation, but for now I’m thinking to just use “specified netspeed is over
X” as the flag indicating that you don’t mind extra queries from outside
your country (or maybe even region).

This is a good idea. After my little AWS instance in Brazil was being
crippled by load, a burlier box I had outside of Brazil was manually put in
the .br zone, and has been able to soak up about 1,800 queries/second.
Resource usage is still relatively low, so I'd be happy to help with some
other zones, having seen how frustrating it can be to be in an underserved
zone.

Estimating load may be trickier than it seems. The AWS instance is at 10
Mbps and averages 2,800 qps. The other box is set to 100 Mbps but only gets
1,800/second. I confess that I don't entirely understand why this is.

-- 
Matt
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to