efficient trouble

Hi

I have some trouble with efficient of my MSSQL 2000, SP4. Server reads about
10-80 MB per secound from phisical disk (average about 20MB), but transfer
into LAN only about 50-500 KB per secound (average 150 KB). It starts about
2-3 month ago and incrise, it means 3 monts ago server read about 2-3MB per
secound, 2 month ago 5MB per secount and so on. Reads is so big, that raid
now can't answer to sql, and users need to wait for result of simple query

I have stoped all jobs and main application, and generally nothing has
chenged (transfer and read was minimally slower). So I think I have:
- unoptimal small aplication or raport or enduser which generate in QA such
big reads,
- or some cache buffer which is filling by server without sense
- or some rootkit with job "to make trouble"

How to find what is going on?

Marek