[dns-operations] cache and subnet option

[Ralf Weber]

Moin!

On 20 Apr 2017, at 3:59, Florian Weimer wrote:

> On 04/20/2017 09:34 AM, Randy Bush wrote:
>>>> [ to my naïve eye, issuing a second query would seem to have, 
>>>> shall
>>>>    we say, interesting scaling properties ]
>>> It does and that is why it usually better to get the distribution
>>> effect by placing resolvers near the edge of the network.
>>
>> is that not the same increase in global query load, just eminating 
>> from
>> more resolvers?
>
> It's worse because those resolvers would have separate caches for 
> infrastructure DNS records.  I don't think those are treated as 
> subnet-dependent by a single, shared cache.
It's the same if you just move resolvers. I was not talking about adding 
resolvers or clients doing recursion themselves. A resolver (at an ISP 
as this is mostly where I work and know stuff) has a certain throughput 
of X qps and serves Y customers. This can be done in a central data 
center or near to the point where subscribers hit the network. Both 
architectures have their merits, but for serving widely dispersed CDNs 
the latter is easier to handle as it doesn't need client subnet.

So long
-Ralf