数据库中存储多个主键； sql vs nosql

Question

I'm developing an app that needs a database with more than one primary key. Its architecture would consist in something like that:

PK == Primary Key

         | column1 | column2(PK) | columns3(PK) | column4(PK) | column5(PK)
---------------------------------------------------------------------------
row1     | word    | string      | string       | string      |  string
---------------------------------------------------------------------------
row2     | word    | string      | string       | string      |  string
---------------------------------------------------------------------------
row3     | word    | string      | string       | string      |  string
---------------------------------------------------------------------------
row4     | word    | string      | string       | string      |  string
---------------------------------------------------------------------------
row5     | word    | string      | string       | string      |  string
---------------------------------------------------------------------------
rowN     | word    | string      | string       | string      |  string

The app will ask the database for the Word (in column1) that correspond to the string specified. The query would be something like that:

SELECT word 
FROM table 
WHERE columnN = 'string' 
LIMIT 1;

where word is the word on column1, columnN is the column 1,2,3,4 or 5 and string is a string that could be contained on columnN

The database will contain million of rows so that I need more that one primary key in order to get the word that correspond to the string very fast.

I've created a database with this structure (but only one primary key on column2 ) and 3,000,000 rows. When I search the word that correspond to the string contained on the column with the primary key, the DB responds me in less than half-second but if I search a word that correspond to a string contained on a column without primary key, the DB responds me on more than half-hour (unacceptable)

So my question is:

How should I structure the database in order to get the response fast independently of the column that contains the string?

I've been reading about noSql databases but I don't know if that's what I need because the database I need is fully relational

Edit: It must be free solution and I have to be able to run it on a domestic computer

Answer 1

I think that adding too many primary keys is not your solution here. Notice that normally primary keys have a size limitation, so you will bump into a road block sooner than later.

What you might be needing here is either a non-clustered index or a vendor-specific technology such as Full-Text Search in MS SQL Server which creates (IMO and by experience) very efficient indexes