Homework #18 Solutions
cs349 -- Networks

Chapter 9

1) ARP and DNS both depend on caches; ARP cache entry lifetimes are
   typically 10 minutes while DNS cache is on the order of days.
   Justify this difference.  What undersirable conseqeunces might
   there be in having too long a DNS cache entry lifetime?

As usual, the lifetime of a cache entry depends on (1) the cost of
recovering if an entry is absent, (2) the cost of using the wrong
entry if it is present, and (3) how often things change.

For ARP, a cache miss requires a local broadcast.  For DNS it requires
multiple connections with DNS servers.  This suggests longer lifetimes
for DNS entries.

If you use an invalid ARP entry, you might send a packet to the
wrong host, but the error would be detected by the higher-level
protocol (e.g. IP).  If you use an invalid DNS entry you might end
up talking to the wrong machine and you would never know.  This
suggests _shorter_ lifetimes for DNS entries.

The frequency of changes in the two cases is probably similar.

In summary, I think the order of days is much too long for DNS
entries, given the high cost of using an obsolete entry.


11) What DNS cache issues are involved in changing the IP address of,
say, a Web server host name?  How might these be minimized?

This is an example where using an obsolete entry can be a serious
problem, since you might get served the wrong page if you contact
the "old" owner of a given name.  This problem might be minimized
by providing a mechanism for sending "DNS update" messages to
inform hosts that their entries have gone bad.