I am looking for a cost-effective solution for time-series database
application that can be programmatically accessed by VB.NET.

DATABASE -  The database will store 1 minute data (real number) for
1000 tags (e.g. tag0001 to tag1000) for 2 months. Data compression
will be a desirable feature for the time-series database.

PROGRAMMING FEATURES - I should be able to calculate a new tag e.g.
tag0999 just by writing a simple expression in VB like
if tag001>tag002, then tag0999 = tag001-tag002 else tag0999=tag003+tag004.

TRENDING - An excellent trending feature including for calculated tags
is a must. I donot want to store a calculation just for trending i.e.
I should be able to trend tag001+tag002 without storing this sum in a
new tag first.

WEB BASED - A web-based deployment of trends is highly desirable.

All of the above can be done by www.osisoft.com's PI software, but I
am looking for a much cheaper solution. One example would be (MySQL +
Dundas Charts + VB.NET) but I am not sure if this can be done.

Can someone please advise on a workable and cheap solution. 

psmsis

Since you broached it, MySQL would be a cheap way of handling it... 
With any database, you'd have to use an INSERT query to do this:
==========QUOTE==========
PROGRAMMING FEATURES - I should be able to calculate a new tag e.g.
tag0999 just by writing a simple expression in VB like
if tag001>tag002, then tag0999 = tag001-tag002 else tag0999=tag003+tag004.
======END QUOTE==========
whatever you get for tag0999, use the query to database: 
INSERT INTO tblname (tagcolumn, tagvalue) VALUES (tagname, tagvalue);

(OR prepopulate with 1000 tags and use an UPDATE to update the appropriate one...)

This is a free comment.

Hi
I am now leaning towards using MATLAB for this application instead of
MySQL. All other things remain the same. Could somone advise how to
accomplish this.

Thanks

psmsis-ga