为什么执行SQL查询时我的静态API如此慢？ 但是当我只是运行查询时它很快

Question

what i would like to achieve is normal length (below 1 minute) of execution restful api time, what my technology that i am using for front end is using Angularjs and Java Spring MVC and for the back end is Java Spring MVC Restful API. But it so slow, it takes more than 5 minute for this only one Stored procedure that only return 2 column. I dunno what i do wrong, because in my local computer it just work fine, but on the server it run very slow, i have other query that complex than this one and it execute faster than this query, here is my query looked like :

    [WEB_SP_Get_Summary]

    ALTER PROCEDURE [dbo].[WEB_SP_Get_Summary]
        @Kd_Plg VARCHAR(50)='KLOP001',
        @Kd_Lokasi VARCHAR(3)='WAT'
    AS
    BEGIN

        SELECT CONVERT(VARCHAR(11),TglJthTempo,106) AS TglJthTempo,
        ISNULL(SUM(PHILIP),0) + ISNULL(SUM(KIM),0) + ISNULL(SUM(HOLIC),0) + ISNULL(SUM(OSRAM),0) + ISNULL(SUM(PANASONIC),0) + ISNULL(SUM([MEGAMEN]),0) AS total_next
        FROM 
        dbo.WEB_F_Dashboard_Get_SisaHutang(@Kd_Plg,@Kd_Lokasi)
        WHERE (TglJthTempo>=GETDATE())
        GROUP BY Kd_Plg,TglJthTempo

    END

Here is my table valued function :


ALTER FUNCTION [dbo].[WEB_F_Dashboard_Get_SisaHutang]
(   
    @Kd_Plg VARCHAR(7),
    @Kd_Lokasi VARCHAR(3)
)
RETURNS @SisaHutang TABLE
(
    Kd_Plg VARCHAR(7),
    TglJthTempo DATETIME,
    PHILIP MONEY,
    KIM MONEY,
    HOLIC MONEY,
    OSRAM MONEY,
    PANASONIC MONEY,
    MEGAMEN MONEY
)
AS
BEGIN
    INSERT INTO @SisaHutang(TglJthTempo,Kd_Plg,PHILIP,KIM,HOLIC,OSRAM,PANASONIC,MEGAMEN)
    SELECT Tgl_JatuhTempo, Kd_Plg, PHILIP, KIM, HOLIC, OSRAM, PANASONIC, MEGAMEN FROM 
        (SELECT a.Divisi,SUM(a.Grandtotal-ISNULL(c.TD,0)) AS sisa_faktur
                ,a.Tgl_JatuhTempo,a.Kd_Plg
        FROM ViewGrandtotal a LEFT JOIN (SELECT No_Faktur, SUM(Total_Distribusi) AS TD
        FROM Trx_DetailDistribusi GROUP BY No_faktur) c ON a.No_Faktur=c.No_faktur 
        WHERE a.GrandTotal - ISNULL(c.TD,0) <> 0
        AND a.Cut_Off = 'A' and a.Kd_Trn='J'
        AND a.Kd_Plg = @Kd_Plg
        AND a.kd_lokasi = @Kd_Lokasi
        GROUP BY a.Divisi, a.Kd_Plg,a.Tgl_JatuhTempo) s
        PIVOT
        (SUM(sisa_faktur)
            FOR Divisi IN (PHILIP,KIM,HOLIC,OSRAM,PANASONIC,MEGAMEN)) AS pvt
RETURN 
END

is there something wrong with my query?

and here is how i do request in my java application :

public List<DashboardPiutangAkanDatang> getDashboardPiutangAkanDatang(String Kd_Plg, String Kd_Lokasi) throws DataAccessException {

    Map<String, Object> params = new HashMap<String, Object>();
       params.put("Kd_Plg", Kd_Plg);
       params.put("Kd_Lokasi", Kd_Lokasi);

    ArrayList<DashboardPiutangAkanDatang> dashboard = (ArrayList<DashboardPiutangAkanDatang>) this.namedParameterJdbcTemplate.query
            ("EXEC WEB_SP_Get_Summary :Kd_Plg, :Kd_Lokasi", params, BeanPropertyRowMapper.newInstance(DashboardPiutangAkanDatang.class));  
    return dashboard;

}

Answer 1

(too long for comment)

The code smells:

• unnecessary encapsulation into TVF which may and I'm sure does cause some performance degradation; get rid of TVF, run updated query at prod server and show actual execution plan
• GROUP BY , GROUP BY , GROUP BY ,.. which is very suspicious;
• LEFT JOIN (SELECT No_Faktur, SUM(Total_Distribusi) AS TD FROM Trx_DetailDistribusi GROUP BY No_faktur) will cause entire scan of Trx_DetailDistribusi whereas you're applying several filters to the left table ViewGrandtotal which means OUTER APPLY or even simple LEFT JOIN could perform much better
• ViewGrandtotal sounds suspicious - view and total are hinting that something is going on inside it and may be questioned as well
• WHERE a.GrandTotal - ISNULL(c.TD,0) <> 0 I don't like it very much
• pivoting and collapsing all the columns into single value right after that? what is there pivot for?
• SQL is not JAVA, such encapsulation and code "reusability" are killing it
• compare server and local table structs, list of indexes
• by the way, is db content equal? or localhost db has let's say 10 records whereas on server there are millions?

Answer 2

If same query giving different performance then something is wrong with your execution plan. when you are executing procedure from your Local SSMS, it uses your SSMS SET Properties. ie SET ARITH_ABORT ON, ... which doesn't in API request.

Please read below article and make required changes in your procedure.

https://www.mssqltips.com/sqlservertip/4318/sql-server-stored-procedure-runs-fast-in-ssms-and-slow-in-application/