SQL query taking very long to execute - sql-server

Question

I have two versions of the same query which I'm trying to execute on SQL Server 2008.
Version 1:
select
count(a.ordernumber) as Orders,
sum(b.agentfees) as AgentFees,
sum(a.revenue) as Revenue,
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date) as Date
from
orders a
join
[DC-SQL-V2].FaxFile.dbo.[agent fees] b on a.ordernumber = b.fforder
where
a.closeoutdate >= '2014-01-01'
and b.dtcreated >= '2014-01-01'
and a.closeoutdate < '2015-01-01'
and b.dtcreated < '2015-01-01'
and a.processserving = 1
and a.branchno = '116'
group by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)
order by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)
Version 1 takes around 3 hours to finish which is obviously insane. So I created Version 2 in which I tried to filter the tables first before joining them, but this still takes pretty long (2 hours and still going).
Version 2:
select
count(a.ordernumber) as Orders,
sum(b.agentfees) as AgentFees,
sum(a.revenue) as Revenue,
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date) as Date
from
(select
ordernumber, revenue, closeoutdate, processserving, branchno
from
ORDERS
where
closeoutdate >= '2014-01-01'
and closeoutdate < '2015-01-01'
and branchno = '116'
and processserving = 1) a
join
(select
agentfees, fforder, dtcreated
from
[DC-SQL-V2].FaxFile.dbo.[agent fees]
where
dtcreated >= '2014-01-01'
and dtcreated < '2015-01-01') b on a.ordernumber = b.fforder
group by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)
order by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)
Here are some additional details:
select count(*) from [DC-SQL-V2].FaxFile.dbo.[agent fees]
779531
select count(*) from ORDERS
3466648
Here are the clustered columns for both tables (among others that I'm not selecting):
ORDERS:
BranchNo
Closeoutdate
ProcessServing
AGENT FEES:
Nothing I'm selecting is clustered
Would someone please be able to suggest how I can improve the performance of my queries?

Answer 1

Assuming [DC-SQL-V2] is a remote, linked server, I'd recommend bringing that data local first.
/* Making broad assumptions here. Substitute correct data types */
CREATE TABLE #tmpAgentFees (
agentfees money,
fforder int,
dtcreated datetime
);
INSERT INTO #tmpAgentFees
(agentfees, fforder, dtcreated)
SELECT agentfees, fforder, dtcreated
FROM [DC-SQL-V2].FaxFile.dbo.[agent fees]
WHERE dtcreated >= '2014-01-01'
AND dtcreated < '2015-01-01';
/* Optionally, add an index to perhaps help performance of the join */
CREATE INDEX IX_tmpAgentFees_fforder
ON #tmpAgentFees(fforder)
INCLUDE (agentfees, dtcreated);
select count(a.ordernumber) as Orders, sum(b.agentfees) as AgentFees, sum(a.revenue) as Revenue,
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date) as Date
from orders a
join #tmpAgentFees b
on a.ordernumber = b.fforder
where a.closeoutdate >= '2014-01-01'
and a.closeoutdate < '2015-01-01'
and a.processserving=1
and a.branchno='116'
group by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)
order by
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date)

Answer 2

There are some general hints
use SQL Server Profiler to find the most costly part
use an index on join column fforder
CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date) is repeated again and again, if it is possible create a computer column in your table and store the value and create an index on this new column with include in index option on
if 3 is not possible use drive table for example as below
select * from
(
select col1,col2,..,CAST('1/' + CAST(DATEPART(Month, a.closeoutdate) as varchar) + '/' + CAST(DATEPART(Year, a.closeoutdate) as varchar) as Date) from table1 ) drivedTable
join .....

Answer 3

The costly part is where you are getting detail data from the foreign server and then summarizing it. You should summarize it their and then only bring that back. It is not clear to me how "dtcreated" and "closeoutdate" are related?
Below I summarized the data by month and then brought it over to the local server. So only 12 rows are coming over instead of one per order. This may not be what you want if the fee should be counted in the "closeoutdate" month. But your code had a problem where it would never count an order or fee if there dates were in different years. (Need to clarify what is needed with that.)
select
a.CloseMonth,
a.Orders,
b.AgentFees
a.Revenue
from
(select year(closeoutdate) * 100 + month(closeoutdate) as CloseMonth,
count(a.ordernumber) as Orders,
sum(a.revenue) as Revenue
from orders
where closeoutdate >= '2014-01-01'
and closeoutdate < '2015-01-01'
and branchno = '116'
and processserving = 1
group by year(closeoutdate) * 100 + month(closeoutdate)) as a
join
(select year(dtcreated) * 100 + month(dtcreated) as FeeMonth,
sum(agentfees) AgentFees
from [DC-SQL-V2].FaxFile.dbo.[agent fees]
where
dtcreated >= '2014-01-01'
and dtcreated < '2015-01-01'
group by year(dtcreated) * 100 + month(dtcreated)) b
on a.CloseMonth = b.FeeMonth
order by a.CloseMonth