bind 9 memory usage

Hi,

I am running bind 9.2.3 on an old 450Mhz with 64Mo of RAM
This server holds few hundreds of zones (~500), and should be queried by approximatively the same
number of users (~500)

It seems that the named process often gets in swread status, and that the dns service regularly
doesn't respond for small periods of time.
here is an output of top command:
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
84 named -22 0 70364K 36884K swread 148:17 0.00% 0.00% named

I don't have any other service running on this machine.
Does that behavior seems normal to you, regarding the machine capabilities, or should I look to my
configuration ?
I thought I had read that dns servers where not "greedy", so I thought that such a machine could be
used as a dns server...
What do you think about that?
Thanks in advance.

Stephane