Database Management Systems

Transcription

1 Database Management Systems Practice #1 Oracle Optimizer Practice bjective Generate the executin plan fr sme SQL statements analyzing the fllwing issues: 1. access paths 2. jin rders and jin methds 3. peratin rders 4. explitatin f indexes defined by the user. The evaluatin will be perfrmed using Oracle Database 10g Express Editin (Oracle XE). Database schema The database cnsists f 3 tables: (EMP, DEPT e SALGRADE). The table schema and sme recrds are shwn in the fllwing. Table EMP EMPNO ENAME JOB MGR HIREDATE SAL COMM DEPT NO 1 COZZA MARIA PROFESSOR 0 09-JUN ECO LUIGI PHDSTUDENT 0 09-JUN CORONA CLARA PHDSTUDENT 2 09-JUN Table DEPT Table SALGRADE DEPTNO DNAME LOC GRADE LOSAL HISAL 1 INFORMATION BARI CHAIRMANSHIP FOGGIA ENVIRONMENT BRINDISI PHYSICS FOGGIA Preliminary steps t perfrm the practice: Cnnectin t the database Open the Oracle SQL Develper prgram (frm Start Menu-All prgrams) Select the Java SDK path Click n the green plus bttn n the left t create a new cnnectin

2 Lgin T lgn thrugh the Web interface, yu have t insert the fllwing parameters: Nme utente (username): bdati[chse a number between 1-100] Passwrd: rac[chse a number between 1-100] Nme hst (hst name): Prt: 1521 SID: xe Fr example, if yu are wrking n pc number 23, the crrespnding username is bdati23 and the passwrd is rac23. Available materials Sme scripts with SQL statements are available t perfrm the fllwing peratins: 1. create an index n a table clumn 2. cmpute statistics fr the database The scripts are available at the curse website in the Scripts.zip archive The scripts can be laded clicking n Open in the File Menu and selecting the.sql file. T execute the script click n the Esegui Script buttn as shwn in the fllwing figure. T view the index statistics, execute the script shw_indexes.sql (r cpy the script cntent and paste it as SQL cmmand)

3 Setting up the ptimizer envirnment At the beginning f wrking sessin yu need t perfrm the fllwing steps: 1. cmpute statistics n tables by means f the Web Interface r by the fllwing script cmp_statistics_tables.sql 2. check if there exist secndary indexes by means f the fllwing SQL query select INDEX_NAME frm USER_INDEXES; if secndary indexes (withut cnsidering system indexes, e.g., SYS_#) have been created, please, drp them by means f the fllwing SQL statement DROP INDEX IndexName; Executin plan cmputatin fr a query T btain the executin plan fr a query thrugh Web interface, it is necessary t execute the query and then t click the Pian di esecuzine buttn (as shwn in Fig.1) t display the executin plan f the query. Fig.1 Useful SQL statements T view the table schema with all attributes: DESCRIBE TableName; T create an index: CREATE INDEX IndexName ON TableName(ClumnName); T cmpute statistics related t indexes: ANALYZE INDEX IndexName COMPUTE STATISTICS; T remve an index: DROP INDEX IndexName; T view the indexes related t a table: SELECT INDEX_NAME FROM USER_INDEXES WHERE table_name='table Name needs t be written in capital letters'; Display statistics related t indexes: SELECT USER_INDEXES.INDEX_NAME as INDEX_NAME, INDEX_TYPE, USER_INDEXES.TABLE_NAME, COLUMN_NAME '(' COLUMN_POSITION ')' as COLUMN_NAME, BLEVEL, LEAF_BLOCKS, DISTINCT_KEYS, AVG_LEAF_BLOCKS_PER_KEY, AVG_DATA_BLOCKS_PER_KEY, CLUSTERING_FACTOR FROM user_indexes, user_ind_clumns WHERE user_indexes.index_name=user_ind_clumns.index_name and user_indexes.table_name=user_ind_clumns.table_name; Display statistics related t tables: SELECT TABLE_NAME, NUM_ROWS, BLOCKS, EMPTY_BLOCKS, AVG_SPACE, CHAIN_CNT, AVG_ROW_LEN FROM USER_TABLES; Display statistics related t table clumns: SELECT COLUMN_NAME, NUM_DISTINCT, NUM_NULLS, NUM_BUCKETS, DENSITY FROM USER_TAB_COL_STATISTICS WHERE TABLE_NAME = 'TableName' ORDER BY COLUMN_NAME; - 3 -

4 Display histgrams: SELECT * FROM USER_HISTOGRAMS; Queries The fllwing queries shuld be analyzed during the practice perfrming the fllwing steps: 1. algebraic expressin represented like a tree structure f the query 2. executin plan selected by Oracle ptimizer when n physical secndary structures are defined 3. Only fr queries frm #4 t #7, Select ne r mre secndary physical structures t increase query perfrmance. Resume f table structures EMP ( EMPNO, ENAME, JOB, MGR, HIREDATE, SAL, COMM, DEPTNO ) DEPT ( DEPTNO, DNAME, LOC ) SALGRADE ( GRADE, LOSAL, HISAL ) Query #1 SELECT * FROM emp, dept WHERE emp.deptn = dept.deptn AND emp.jb = 'ENGINEER'; Change the ptimizer gal frm ALL ROWS (best thrughput) t FIRST_ROWS (best respnse time) by means f the fllwing hint. Set different values fr n. SELECT /*+ FIRST_ROWS(n) */ * FROM emp, dept WHERE emp.deptn = dept.deptn AND emp.jb = 'ENGINEER'; Query #2 Disable the hash jin methd by means f the fllwing hint: (/*+ NO_USE_HASH(e d) */) SELECT /*+ NO_USE_HASH(e d) */ d.deptn, AVG(e.sal) FROM emp e, dept d WHERE d.deptn = e.deptn GROUP BY d.deptn; Query #3 Disable the hash jin methd by means f the fllwing hint: (/*+ NO_USE_HASH(e d) */) SELECT /*+ NO_USE_HASH(e d) */ ename, jb, sal, dname FROM emp e, dept d WHERE e.deptn = d.deptn AND NOT EXISTS (SELECT * FROM salgrade WHERE e.sal = hisal); Queries #4 Select ne r mre secndary structures t ptimize the fllwing query: select avg(e.sal) frm emp e where e.deptn < 10 and e.sal > 100 and e.sal < 200; Cmpare query perfrmance using distinct secndary structures n different attributes with the ne achieved by a unique secndary structure n multiple attributes

5 Query #5 Select ne r mre secndary structures t ptimize the fllwing query: select dname frm dept where deptn in (select deptn frm emp where jb = 'PHILOSOPHER'); Query #6 Select ne r mre secndary structures t ptimize the fllwing query (remve already existing indexes t cmpare query perfrmance with and withut indexes): select e1.ename, e1.empn, e1.sal, e2.ename, e2.empn, e2.sal frm emp e1, emp e2 where e1.ename <> e2.ename and e1.sal < e2.sal and e1.jb = 'PHILOSOPHER' and e2.jb = 'ENGINEER'; Query #7 Select ne r mre secndary structures t ptimize the fllwing query: select * frm emp e, dept d where e.deptn = d.deptn and e.sal nt in (select hisal frm salgrade where hisal > 500 and hisal < 1900); - 5 -