Help analyzing a single query perfomance

Question

Hi,
we're using MaxDB 7.7 in a large e-commerce project.
Recently we're having performances issues and I'm trying to understand what's the problem.
This is the query
select articolo2_.ID as col_0_0_, articolo2_.CODICE as col_1_0_, articolo2_.CODICEREPOSITORY as col_2_0_, articolo2_.DESCRIZIONE as col_3_0_, marchio5_.NOME as col_4_0_, articolo2_.MODELLO as col_5_0_, categoria6_.NOME as col_6_0_, unitamisur7_.DESCRIZIONE as col_7_0_, articoloco1_.ID as col_8_0_, articoloco1_.CODICE as col_9_0_, articoloco1_.STATOCOMMERCIALE as col_10_0_, articoloco1_.DATA_STATOCOMMERCIALE as col_11_0_, articolo2_.ALTRO1 as col_12_0_, articolo2_.ALTRO2 as col_13_0_, articolo2_.ALTRO3 as col_14_0_, articolo2_.ALTRO4 as col_15_0_, articolo2_.ALTRO5 as col_16_0_, articoloco1_.ATTIVO as col_17_0_, articoloco1_.LOTTOORDINE as col_18_0_, articoloco1_.MINIMOORDINE as col_19_0_, articoloco1_.SCORTAMINIMA as col_20_0_, MIN(prezzo0_.PREZZO) as col_21_0_, SUM(dispmagazz4_.DISPONIBILITA) as col_22_0_, SUM(dispmagazz4_.DISPFUTURA) as col_23_0_, iva8_.CODICE as col_24_0_, iva8_.PERC as col_25_0_, articolo2_.STIMAPESO as col_26_0_, MAX(prezzo0_.FLAG) as col_27_0_, articolo2_.COMPOSTO as col_28_0_, articoloco1_.ELEMENTO_ARTICOLOCOMPOSTO as col_29_0_, articoloco1_.DATA_CREAZIONE as col_30_0_, articoloco1_.DATA_ULTIMAMODIFICA as col_31_0_, MIN(prezzo0_.PREZZOPREC) as col_32_0_, MIN(prezzo0_.DATAPREZZOPREC) as col_33_0_, MIN(prezzo0_.PREZZOPRECNOFLAG) as col_34_0_, articolo2_.ALTEZ as col_35_0_, articolo2_.LARG as col_36_0_, articolo2_.PROF as col_37_0_, articolo2_.TAGLIA as col_38_0_, articolo2_.COLORE as col_39_0_, articolo2_.TIPOMISURE as col_40_0_, articolo2_.STIMACOLLI as col_41_0_, articolo2_.KEYWORDS as col_42_0_, MIN(prezzo0_.PREZZOIVATO) as col_43_0_, MIN(prezzo0_.PREZZOPRECIVATO) as col_44_0_, MIN(prezzo0_.PREZZOPRECNOFLAGIVATO) as col_45_0_, articolo2_.CODSTRUTTURAVARIANTE as col_46_0_, articolo2_.VARIANTE1 as col_47_0_, articolo2_.VARIANTE2 as col_48_0_, articolo2_.VARIANTE3 as col_49_0_, articolo2_.VARIANTE4 as col_50_0_, articolo2_.VARIANTE5 as col_51_0_, articolo2_.VARIANTE6 as col_52_0_, articolo2_.VARIANTE7 as col_53_0_, articolo2_.VARIANTE8 as col_54_0_, articolo2_.CARATTERISTICA1 as col_55_0_, articolo2_.CARATTERISTICA2 as col_56_0_, articolo2_.CARATTERISTICA3 as col_57_0_, articolo2_.CARATTERISTICA4 as col_58_0_, articolo2_.CARATTERISTICA5 as col_59_0_, MIN(articolifo3_.ID_PUNTOPARTENZA) as col_60_0_, iva10_.CODICE as col_61_0_, iva10_.PERC as col_62_0_, articolo2_.QTACONFEZIONE as col_63_0_, articolo2_.QTACARTONE as col_64_0_, articolo2_.QTABANCALE as col_65_0_
from ECF3.PREZZO prezzo0_, ECF3.ARTICOLO_COMMERCIALE articoloco1_, ECF3.ARTICOLO articolo2_, ECF3.ARTICOLO_FORNITORE articolifo3_, ECF3.MARCHIO marchio5_, ECF3.CATEGORIA categoria6_, ECF3.UNITAMISURA unitamisur7_, ECF3.IVA iva8_, ECF3.IVA iva10_, ECF3.DISPMAGAZZINO dispmagazz4_, ECF3.MACROCATEGORIA macrocateg13_, ECF3.REPARTO reparto14_
where prezzo0_.ID_ARTICOLOCOMM=articoloco1_.ID and articoloco1_.ID_ARTICOLO=articolo2_.ID and articolo2_.ID=articolifo3_.ID_ARTICOLO(+) and articolo2_.ID_MARCHIO=marchio5_.ID and articolo2_.ID_CATEGORIA=categoria6_.ID and articolo2_.ID_UM=unitamisur7_.ID and articolo2_.ID_IVA=iva8_.ID and articolo2_.ID_IVAINGROSSO=iva10_.ID and categoria6_.ID_MACROCATEGORIA=macrocateg13_.ID and macrocateg13_.ID_REPARTO=reparto14_.ID and articolifo3_.ABITUALE='S' and dispmagazz4_.ID_ARTICOLOCOMM=articoloco1_.ID and articoloco1_.ID_AZIENDA=1 and articoloco1_.ID_CANALE=9 and articoloco1_.ATTIVO='A' and prezzo0_.ID_LISTINO=47 and prezzo0_.PREZZO>0 and reparto14_.NOME='ELDOM' and (upper(articolo2_.DESCRIZIONE) like '%mp3%' or upper(articolo2_.MODELLO) like '%mp3%' or upper(articolo2_.KEYWORDS) like '%mp3%')
group by articolo2_.ID , articolo2_.CODICE , articolo2_.CODICEREPOSITORY , articolo2_.DESCRIZIONE , marchio5_.NOME , articolo2_.MODELLO , categoria6_.NOME , unitamisur7_.DESCRIZIONE , articoloco1_.ID , articoloco1_.CODICE , articoloco1_.STATOCOMMERCIALE , articoloco1_.DATA_STATOCOMMERCIALE , articolo2_.ALTRO1 , articolo2_.ALTRO2 , articolo2_.ALTRO3 , articolo2_.ALTRO4 , articolo2_.ALTRO5 , articoloco1_.ATTIVO , articoloco1_.LOTTOORDINE , articoloco1_.MINIMOORDINE , articoloco1_.SCORTAMINIMA , iva8_.CODICE , iva8_.PERC , articolo2_.STIMAPESO , articolo2_.COMPOSTO , articoloco1_.ELEMENTO_ARTICOLOCOMPOSTO , articoloco1_.DATA_CREAZIONE , articoloco1_.DATA_ULTIMAMODIFICA , articolo2_.ALTEZ , articolo2_.LARG , articolo2_.PROF , articolo2_.TAGLIA , articolo2_.COLORE , articolo2_.TIPOMISURE , articolo2_.STIMACOLLI , articolo2_.KEYWORDS , articolo2_.CODSTRUTTURAVARIANTE , articolo2_.VARIANTE1 , articolo2_.VARIANTE2 , articolo2_.VARIANTE3 , articolo2_.VARIANTE4 , articolo2_.VARIANTE5 , articolo2_.VARIANTE6 , articolo2_.VARIANTE7 , articolo2_.VARIANTE8 , articolo2_.CARATTERISTICA1 , articolo2_.CARATTERISTICA2 , articolo2_.CARATTERISTICA3 , articolo2_.CARATTERISTICA4 , articolo2_.CARATTERISTICA5 , iva10_.CODICE , iva10_.PERC , articolo2_.QTACONFEZIONE , articolo2_.QTACARTONE , articolo2_.QTABANCALE
order by MIN(prezzo0_.PREZZO) ASC
and this is the explain result
SCHEMANAME    TABLENAME           COLUMN_OR_INDEX                     STRATEGY                                  PAGECOUNT
              UNITAMISUR7_                                            TABLE SCAN                                         1
              PREZZO0_            PREZZO_LISTINO_idx                  JOIN VIA INDEXED COLUMN                        13655
                                  ID_LISTINO                               (USED INDEX COLUMN)                           
              ARTICOLOCO1_        ID                                  JOIN VIA KEY COLUMN                             5401
              REPARTO14_          IDX_NOME_MONDO                      JOIN VIA INDEXED COLUMN                            1
                                                                      TABLE HASHED                                       
                                  NOME                                     (USED INDEX COLUMN)                           
              ARTICOLIFO3_        IDX_ARTICOLO_FORNITORE_ARTICOLO     JOIN VIA INDEXED COLUMN                         5478
                                  ID_ARTICOLO                              (USED INDEX COLUMN)                           
              ARTICOLO2_          ID                                  JOIN VIA KEY COLUMN                             8098
              MARCHIO5_           ID                                  JOIN VIA KEY COLUMN                                5
                                                                      TABLE HASHED                                       
              CATEGORIA6_         ID                                  JOIN VIA KEY COLUMN                                8
                                                                      TABLE HASHED                                       
              IVA8_               ID                                  JOIN VIA KEY COLUMN                                1
                                                                      TABLE HASHED                                       
              IVA10_              ID                                  JOIN VIA KEY COLUMN                                1
                                                                      TABLE HASHED                                       
              DISPMAGAZZ4_        DISPMAGAZZINO_IDARTICOLOCOMM_IDX    JOIN VIA INDEXED COLUMN                          801
                                  ID_ARTICOLOCOMM                          (USED INDEX COLUMN)                           
              MACROCATEG13_       ID                                  JOIN VIA KEY COLUMN                                1
                                                                      TABLE HASHED                                                                               
NO TEMPORARY RESULTS CREATED                  
INTERNAL      TEMPORARY RESULT                                        TABLE SCAN                                         1
              JDBC_CURSOR_54                                               RESULT IS COPIED   , COSTVALUE IS        289903
The query takes an average of 25 seconds in working hours. It returns 611 rows. MaxDB 7.7 running on Linux, 2 processors, 4 cores each, 32GB Ram. Cache size is 4GB for this db and data size is about 5GB. Cache hit 100%.
What I'm not sure is if the query is already optimized and the problem is the workload on the server or the query can be optimized.
The query is generated by Hibernate so I really can't tweak the SQL.
Thank you for any suggestion !

Answer 1

Dear Lars,
thanks for your help.
I think the problem is on some joins because removing them and using an added field in place of them (de-normalyzing information) the query is much faster.
First, all the joins are backed by indexes with integer keys.
These are the tables involved  (i removed the fields not related to joins for clarity), query time are not average, they're a single run times in working hours :
CREATE TABLE ECF3.ARTICOLO (
     ID INTEGER NOT NULL,
     DESCRIZIONE VARCHAR() ASCII(512) NOT NULL,
     MODELLO VARCHAR() ASCII(100) NOT NULL,
     KEYWORDS VARCHAR() ASCII(512),
     ID_UM INTEGER,
     ID_IVA INTEGER,
     ID_IVAINGROSSO INTEGER,
     ID_CATEGORIA INTEGER,
     ID_MARCHIO INTEGER,
     PRIMARY KEY (ID)
The table contains 115.729 rows, 4.857 rows match the search condition on this table fields, like on '%MP3%' (query runs in 0.4 sec).
CREATE TABLE ECF3.ARTICOLO_COMMERCIALE (
     ID INTEGER NOT NULL,
     ID_AZIENDA INTEGER,
     ID_CANALE INTEGER,
     ID_ARTICOLO INTEGER,
     ATTIVO CHAR() ASCII(1) NOT NULL,
     PRIMARY KEY (ID)
The table contains 413.916 rows, 45.086 rows match the search condition on this table field, ID_AZIENDA=1, ID_CANALE=9, ATTIVO='A' (query runs in 0.48 sec).
CREATE TABLE ECF3.ARTICOLO_FORNITORE (
     ID INTEGER NOT NULL,
     ABITUALE CHAR() ASCII(1),
     ID_ARTICOLO INTEGER,
     PRIMARY KEY (ID)
The table contains 115.503 rows, all match the search condition on this table field, ABITUALE='S' (query runs in 0.7 sec).
CREATE TABLE ECF3.PREZZO (
     ID INTEGER NOT NULL,
     PREZZO FIXED(14,3),
     ID_LISTINO INTEGER,
     ID_ARTICOLOCOMM INTEGER,
     PRIMARY KEY (ID)
The table contains 2.246.518 rows, 58.396 match the search condition on this table field, ID_LISTINO=47 AND PREZZO>0 (query runs in 0.4 sec)
CREATE TABLE ECF3.DISPMAGAZZINO (
     ID_MAGAZZINO INTEGER NOT NULL,
     ID_ARTICOLOCOMM INTEGER NOT NULL,
     PRIMARY KEY (ID_MAGAZZINO,ID_ARTICOLOCOMM)
Table contains 404.664 rows, no conditions on query, only joins with other table ARTICOLO_COMMERCIALE.
There are then the smaller tables
CREATE TABLE ECF3.REPARTO (
     ID INTEGER NOT NULL,
     NOME VARCHAR() ASCII(100) NOT NULL,
     PRIMARY KEY (ID)
CREATE TABLE ECF3.MACROCATEGORIA (
     ID INTEGER NOT NULL,
     NOME VARCHAR() ASCII(100) NOT NULL,
     ID_REPARTO INTEGER DEFAULT           0 NOT NULL,
     PRIMARY KEY (ID)
CREATE TABLE ECF3.CATEGORIA (
     ID INTEGER NOT NULL,
     ID_MACROCATEGORIA INTEGER NOT NULL,
     NOME VARCHAR() ASCII(100) NOT NULL,
     PRIMARY KEY (ID)
This is a 'tree', reparto has only 5 rows, macrocategoria has 29 rows, categoria has 1120 rows. The search condition on reparto gets to 765 categoria rows.
CREATE TABLE ECF3.MARCHIO (
     ID INTEGER NOT NULL,
     NOME VARCHAR() ASCII(100) NOT NULL,
     PRIMARY KEY (ID)
The table contains 1189 rows
CREATE TABLE ECF3.IVA (
     ID INTEGER NOT NULL,
     PRIMARY KEY (ID)
This table contains 15 rows.
CREATE TABLE ECF3.UNITAMISURA (
     ID INTEGER NOT NULL,
     PRIMARY KEY (ID)
This table contains 6 rows.
The CATEGORIA, MARCHIO, MACROCATEGORIA, REPARTO was added later to the structure. Before this we had a field CATEGORIA, MARCHIO, REPARTO in the ARTICOLO table.
Using the fields instead of the joins the query is fast (2 sec.). Adding these tables has slowed down to 25 sec. (15 in non working hours).
This is what I don't understand, they're small tables, I understand the joins add complexity but the time difference between the two version of the model is very high !
Thanks for help and sorry for the long post.