[dns-operations] Peak qps/subscriber rule of thumb for residential ISP networks

[bert hubert]

On Sat, Dec 06, 2014 at 08:30:42PM -0600, Frank Bulk wrote:
> What's a typical peak qps/subscriber in a residential ISP network?  We have
> 6 recursing resolvers in our network (some behind a load-balancer), and if I
> add up each of their peak qps and then divide that total by our subscriber
> count I get 0.042 qps/subscriber.  I understand that some customer are going

We've long used 0.1 qps/subscriber, which is not too dissimilar from your
number. 0.1 was based on observations from a variety of large scale powerdns
users, and it still appears to hold.

Important factors that changed in the 5 years we've been using this number
include: the explosion of devices in households, browser prefetching, CPE
behaviour (many assign themselves as DNS for the house), and of course some
8.8.8.8 substitution.

But it appears you won't be too far wrong with 0.1.

	Bert
-- 
PowerDNS Website: http://www.powerdns.com/
Contact us by phone on +31-15-7850372