最近很長一段時間都在優化一個項目，這個項目存在許多問題，從數據架構到工作流程，我一直在思考一些有關Oracle性能優化的問題，我跟開發實施人員進行過多次交流，發現存在許多交流障礙，許多問題實施人員覺得sql語句執行很快，并不存在性能問題，我花了大量時間要他看執行計劃，說明當數據積累到一定時間后，會執行越來越慢。

我舉一個例子來說明問題：

SELECT v2.*

FROM consultationrecorddoctor t1

RIGHT OUTER JOIN

(SELECT cr.*, crc.modifydate, crc.modifyuserid, crc.consultationtime,

crc.applyconsultationdeptid, crc.askconsultationdeptid,

crc.consultationdeptid, crc.casehistory_right,

crc.consultationorder_right, crc.consultationidea_right,

crc.otherhospital_right, crc.consultationdate,

crc.maindoctorid, c1.deptname_vchr AS applydeptname,

c2.deptname_vchr AS askdeptname,

c3.deptname_vchr AS deptname,

f_getempnamebyno (crc.maindoctorid) AS maindocname

FROM consultationrecord cr,

consultationrecordcontent crc,

t_bse_deptdesc c1,

t_bse_deptdesc c2,

t_bse_deptdesc c3,

(SELECT cr.inpatientid, cr.inpatientdate, cr.opendate,

cr.createdate,

MAX (crc.modifydate) AS maxmodifydate

FROM consultationrecord cr,

consultationrecordcontent crc

WHERE (crc.applyconsultationdeptid = '0000208')

AND cr.inpatientid = crc.inpatientid

AND cr.inpatientdate = crc.inpatientdate

AND cr.opendate = crc.opendate

AND cr.status = 0

GROUP BY cr.inpatientid,

cr.inpatientdate,

cr.opendate,

cr.createdate) v1

WHERE (crc.applyconsultationdeptid = '0000208')

AND cr.inpatientid = crc.inpatientid

AND cr.inpatientdate = crc.inpatientdate

AND cr.opendate = crc.opendate

AND cr.status = 0

AND crc.applyconsultationdeptid = c1.deptid_chr

AND crc.askconsultationdeptid = c2.deptid_chr

AND crc.consultationdeptid = c3.deptid_chr

AND cr.inpatientid = v1.inpatientid

AND cr.inpatientdate = v1.inpatientdate

AND cr.opendate = v1.opendate

AND v1.maxmodifydate = crc.modifydate

AND crc.maindoctorid IS NOT NULL) v2

ON t1.inpatientid = v2.inpatientid

AND t1.inpatientdate = v2.inpatientdate

AND t1.opendate = v2.opendate

AND t1.modifydate = v2.modifydate

AND t1.employeeflag = 1

WHERE t1.employeeid IS NULL

ORDER BY t1.employeeid DESC, v2.askdeptname, v2.consultationdate;

查詢使用了一個右連接，查詢的表并沒有包含t1表的字段，而且在實際的t1.employeeid中

根本不存在NULL的值，實際上是從一個大的結果集剔除一個大結果集的而獲得一個小的結

果集，隨著crc表數據量加大，會越來越慢！