Question

I have a written a query in which I create a string and take distinct values from the table based on conditions.

The table has some 5000 rows. This query takes almost 20 second to execute.

I believe that the string comparisons have made the query so slow. But I wonder what are my alternatives.

Query:

select distinct
   Convert(nvarchar, p1.trafficSerial) +' ('+  p1.sourceTraffic + ' - ' + p1.sinkTraffic + ' )' as traffic
from 
   portList as p1
inner join 
   portList as p2 ON p1.pmId = p2.sinkId
                  AND P1.trafficSerial IS NOT NULL 
                  AND (p1.trafficSerial = p2.trafficSerial) 
                  AND (P1.sourceTraffic = P2.sourceTraffic) 
                  AND (P1.sinkTraffic = P2.sinkTraffic)
where 
   p1.siteCodeID = @SiteId

Answer 1

One option is to create a computed column and create an index on that

This article discusses it http://blog.sqlauthority.com/2010/08/22/sql-server-computed-columns-index-and-performance/

ALTER TABLE dbo.portList ADD
traffic AS Convert(nvarchar,trafficSerial) +' ('+  sourceTraffic + ' - ' + sinkTraffic + ' )' PERSISTED
GO

CREATE NONCLUSTERED INDEX IX_portList_traffic
ON dbo.portList  (traffic)
GO

select distinct traffic from dbo.portList 

You should also make sure each of the columns in your join relationships have indexes on them:

p1.trafficSerial & p2.trafficSerial
P1.sourceTraffic & P2.sourceTraffic
P1.sinkTraffic & P2.sinkTraffic

and the column for your filter: p1.siteCodeID