[erlang-questions] very large key lookup

[ke han]

I am providing software architecture advice to a company and have  
pointed them towards erlang in building their custom server.
The server basically does one thing:  take in a 32 byte key from  
client sockets and lookup for a match against an internal key table.   
If a match is found, it checks against the matched key's 128 byte key  
for secondary false-positive elimination.  Sorry I can't provide the  
context of this app as the company is in stealth mode at the moment.   
Hopefully they will be a proud erlang supporter once they launch this  
in a few months.
We will have hundreds of millions of these 32 byte keys in the the  
server's "DB" and each key's associated 128 byte key.
I am looking for advice on the format for these keys along with which  
libraries to use for balance between memory  and lookup performance.   
The current format is 0-9 + a-z character strings.  We can change  
this since the lookup tables and the input from clients does not need  
to be human readable.
thanks, ke han