I have a stored procedure which identifies any changes in candidate information from the last SSIS job run. Then it proceeds to save the changes in a sync table. ``` CREATE PROCEDURE [dbo].[SingleCandidateView_IdentifyCandidateChanges] ( @MoreToProcess BIT = 0 OUTPUT ) AS SET NOCOUNT ON DECLARE @BatchSize INT SELECT @MoreToProcess = 0 , @BatchSize = 200000 DECLARE @Changes TABLE ( RSACandidateNo VARCHAR(10) NOT NULL , CentreNo CHAR(5) NOT NULL , CandidateModifiedDate DATETIME NOT NULL , ChangeStatus CHAR(1) NOT NULL ) -- If this is the first run (or the CandidateSync table is empty for another reason) then just take the hit now -- and process the entire set of candidates in one go. IF EXISTS(SELECT TOP 1 1 FROM CandidateSync) SET ROWCOUNT @BatchSize INSERT INTO @Changes ( RSACandidateNo , CentreNo , CandidateModifiedDate , ChangeStatus ) SELECT COALESCE(RTRIM(Known.RSACandidateNo), CandidateSync.RSACandidateNo) AS RSACandidateNo , COALESCE(Known.CentreNo, CandidateSync.CentreNo) AS CentreNo , COALESCE(Candidate.ModifiedDate, CandidateSync.CandidateModifiedDate) AS CandidateModifiedDate , CASE WHEN CandidateSync.RSACandidateNo IS NULL THEN 'I' WHEN Known.RSACandidateNo IS NULL THEN 'D' ELSE 'U' END AS ChangeStatus FROM Candidate WITH(INDEX(IX_Candidate)) INNER MERGE JOIN ( SELECT Registration.RSACandidateNo , Registration.CentreNo FROM Registration WITH(INDEX(IX_Registration)) UNION SELECT AssessHeaderValid.CandidateNo , AssessHeaderValid.CentreNo FROM AssessHeaderValid WITH(INDEX(IX_AssessHeaderValid)) WHERE AssessHeaderValid.Deleted = 0 UNION SELECT CandAnsHeaderValid.CandidateNo , CandAnsHeaderValid.CentreNo FROM CandAnsHeaderValid WITH(INDEX(IX_CandAnsHeaderValid)) ) AS Known ON Known.RSACandidateNo = Candidate.RSACandidateNo FULL OUTER MERGE JOIN CandidateSync WITH(INDEX(PK_CandidateSync)) ON CandidateSync.RSACandidateNo = Known.RSACandidateNo AND CandidateSync.CentreNo = Known.CentreNo WHERE CandidateSync.RSACandidateNo IS NULL OR Known.RSACandidateNo IS NULL OR CandidateSync.CandidateModifiedDate < Candidate.ModifiedDate OR CandidateSync.ChangeStatus = 'D' -- Always include changes to pending deletions so that the scenario where a record has been resurrected is handled correctly. OPTION ( MERGE UNION ) IF @@ROWCOUNT = @BatchSize SET @MoreToProcess = 1 SET ROWCOUNT 0 INSERT INTO CandidateSync WITH (PAGLOCK) ( RSACandidateNo , CentreNo , CandidateModifiedDate , ChangeStatus ) SELECT Changes.RSACandidateNo , Changes.CentreNo , Changes.CandidateModifiedDate , CASE Changes.ChangeStatus WHEN 'D' THEN 'D' ELSE 'U' END FROM @Changes Changes WHERE Changes.ChangeStatus = 'I' UPDATE CandidateSync WITH (ROWLOCK) SET CandidateModifiedDate = Changes.CandidateModifiedDate , ChangeStatus = CASE Changes.ChangeStatus WHEN 'D' THEN 'D' ELSE 'U' END , ProcessingStatus = 'U' FROM @Changes Changes WHERE CandidateSync.RSACandidateNo = Changes.RSACandidateNo AND CandidateSync.CentreNo = Changes.CentreNo AND CandidateSync.ProcessingStatus = 'S' AND Changes.ChangeStatus <> 'I' ``` The Indexes are setup as: | Name | Type | Keys | |:--------|:---------|:------| |IX_Candidate | nonclustered located on PRIMARY | RSACandidateNo, ModifiedDate| |IX_Registration | nonclustered located on PRIMARY | RSACandidateNo, CentreNo| |IX_AssessHeaderValid | nonclustered located on PRIMARY | CandidateNo, CentreNo, Deleted| |IX_CandAnsHeaderValid | nonclustered located on PRIMARY | CandidateNo, CentreNo| |PK_CandidateSync | clustered, unique, primary key located on PRIMARY| RSACandidateNo, CentreNo| The approximate count of rows in the tables are as: | Table| Count |:--------|:---------| |Candidate | 10M| |Registration | 12M| |AssessHeaderValid | 3.4M| |CandAnsHeaderValid | 8M| |CandidateSync | 8.4M| The actual execution plan: [![enter image description here][1]][1] [1]: https://i.sstatic.net/KcUCm.png [PasteThePlan](https://www.brentozar.com/pastetheplan/?id=HyIks3oG3) The tables Candidate, Registration, CandAnsHeaderValid and AssessHeaderValid are updated twice daily by an SSIS job run, which need to be synced to the CandidateSync table, as it keeps record of the records which have been already processed. The issue is that this query is taking almost a minute to complete and is completing massive table scans with sorts spilling to tempdb and millions of rows being pumped into table variables. Is there any way I can reduce the load on the server?