[Click] TrieIPLookup element and routing cache

[Marko Zec]

On Monday 14 February 2005 10:51, Davide Bolognesi wrote:
> Hi,
>
> I'm Davide Bolognesi, I'm a student at Politecnico di Torino, Italy.
> I'm working on the E.U.R.O project
> (http://www.diit.unict.it/euro/index_en.htm) for my Laurea degree
> thesis.
>
> I've downloaded Click-1.4.3 and I would like to know more about the
> TrieIPLookup element: could you please point me to a paper/book
> describing the algorithm implemented? I've already read the comments
> in the C++ code, and I haven't found any clue!
>
> All *IPLookup elements seem to have only minimal routing cache
> support, is anybody aware of any element that implements a better
> routing cache?

It is questionable how much benefit a routing cache can provide in 
today's Internet.  Consider backbone routers carrying thousands of 
unrelated flows, or scenarios where worms randomly scan address space 
searching for new vulnerable nodes.  Although certainly there are cases 
where routing caches might help, employing a lookup scheme which allows 
for full routing lookups on each packet yields much more predictable 
behavior.  Both TrieIPLookup and DirectIPLookup should be fast enough 
for a PC router; the throughput bottlenecks will probably occur at 
other places, most likely the PCI bus will start choking before the 
CPU...

> What about ARP elements? Do they implement any ARP cache?

I guess any functional ARP implementation would need a cache for storing 
a map of valid IP/MAC pairs (until they get expired by timers), 
otherwise it would generate a constant storm of ARP requests.  The 
cache in the arpquerier element can store up to 256 entries by default.

Hope this helps,

Marko

>
> Thank you in advance for any help,
>
> Cheers,
>
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click