High Scalability Building bigger, faster, more reliable websites.

I don't understand why you cannot use an index. To be blunt, without the use of indexes you cannot search through millions of rows quickly. How do you expect a database engine to do that unless everything is laid out in a structured logical order i.e. accessed using indexes?
If funny or hollywood is a category, then you would normally represent these as categoryid's from a related lookup table. Then you would search the database with the where predicate based on a selected categoryid. With a clustered index on this column this would normally be very fast. In Sql Server 2005, I have a huge indexed view with a similar amount of columns and perform such queries in under a second. Without the use of indexes on the source tables and the view however, the query time can go up to 10 seconds! You cannot query millions of rows without indexes and expect high performance, especially if you have anything other than narrow tables containing only ints.

Thanks for your reply
In column ,data is like "funny ;hollywood ;fast ;youtube;".
For every column i need to compare each and every word contained in that column .If i use substring kind of function then in that case database don't use index.I want to store tags in column and use it for searching.

Is there any other solution like Lucene or Hadoop.I heard Lucene is very good in searching but not sure about its scalability.Is we can use Hadoop for online search or it is only suitable for indexing our data offline.