index skipped in favor of seq scan.

I am working with putting syslog logs into a database, I'm parsing the
logs and using the key information for my fields.  With my test data of
~200K rows the optimizer used my b-tree index that I created for an
oft-used where clause.  When the table grew to over 800K rows the index
was no longer used.  The field in question contains IP addresses, but
uses varchar.  The values are _not_ unique.  One particular address has
150K entries.  How can I keep my where queries speedy?

-ryan