Caching-Only Server

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

Hi,

I wonder if you can help, I currently work in a large organisation that has sites globaly.

In order to reduce WAN traffic we want to implement caching only servers to reduce traffic.

My question is, as the cached entries reside in RAM, what is the best way to implement redundancy for when the WAN link goes down and when the cache is cleared...either by rebooting or harware failure???

Can you also confirm whether it is possible to establish a standard TTL for all A records cached as these seem to vary from anything between 5 and 30 minutes??

I look forward to hearing from you.

Thanks and regards
 
In Jeff_The_Ref <[email protected]> posted a question
Then Kevin replied below:
: Hi,
:
: I wonder if you can help, I currently work in a large organisation
: that has sites globaly.
:
: In order to reduce WAN traffic we want to implement caching only
: servers to reduce traffic.
:
: My question is, as the cached entries reside in RAM, what is the best
: way to implement redundancy for when the WAN link goes down and when
: the cache is cleared...either by rebooting or harware failure???
:
: Can you also confirm whether it is possible to establish a standard
: TTL for all A records cached as these seem to vary from anything
: between 5 and 30 minutes??
:
: I look forward to hearing from you.
:
: Thanks and regards

An MS DNS has a MaxCacheTtl of one day, regardless of what the actual record
has.
MS DNS records have a default Ttl of one hour, no matter what you set the
MaxCacheTtl to, it will not cache for longer than the Ttl of the record.

I don't think that introducing a caching only DNS in the mix will help a
whole lot if the Ttl of the records are still one hour.
Many of these records can probably have Ttls of a lot more then one hour
especially the DC records which are only 10 or 20 Minutes. A caching DNS
will not help these records much because they won't get cached very long
anyway.
You might also consider, setting up a caching DNS if you can delegated as
many of the records or subzones as possible, depending on you namespace
structure this could be very difficult, because you may need to start your
delegations at the Root. It may pay off in the end by reducing network
activity.
 
J> Can you also confirm whether it is possible to establish a
J> standard TTL for all A records cached as these seem to vary
J> from anything between 5 and 30 minutes??

This is not possible (without modifying one's caching proxy DNS server
software to make it behave in unconventional ways). The TTL of a resource
record set is specified _by the publisher_ of that resource record set, not by
a caching proxy DNS server. If you are not the publisher, you do not control
what TTL is specified. The only thing that caching proxy DNS servers commonly
do is place a cap on TTL values, so that they don't cache data for
unreasonably long periods even if the publisher has specified a large TTL
value. (For example: Microsoft's DNS server, by default, places an upper
bound on resource record set TTLs of 1 day. ISC's BIND, by default, my
caching proxy DNS server, by default, and Dan Bernstein's "dnscache" all place
an upper bound on resource record set TTLs of 1 week.)

If you _are_ the publisher, then if you don't like the inconsistent TTLs on
the resource record sets that you are publishing, simply change them.
 
Back
Top