Explain for a very long running query

Question

Dear gurus,
I have an issue with a long running query which takes around 2 hrs to execute. The tables being used here pertains to 11i, so there is no scope of indexing.
SELECT peia.expenditure_item_id, DECODE (peia.system_linkage_function, 'ST', 'TIME', 'EXP' ) AS CATEGORY,
pea.incurred_by_person_id AS employee_id, emp.attribute6 AS q_number, NVL (emp.employee_number, npw_number) AS employee_number, emp.full_name AS employee_name, peia.project_id,
ppa.segment1, ppa.NAME, func_get_proj_attribute (peia.project_id, NULL, 'PROJECT ENTITY NAME' ) AS project_entity,
func_get_proj_attribute (peia.project_id, NULL, 'PROJECT MANAGER') AS project_manager,
DECODE (ppa.distribution_rule, 'WORK/WORK', 'TM', 'WORK/EVENT', 'TM with Milestone' ) AS project_classification,
ppa.project_type AS funding_type, func_get_proj_attribute(peia.project_id, NULL, 'CUSTOMER GROUP' ) AS customer_group,
func_get_proj_attribute(peia.project_id, NULL, 'ACCOUNT MANAGER' ) AS account_manager,
pt.top_task_id, func_get_task_attribute(pt.top_task_id, 'TASK NUMBER' ) AS task_number,
func_get_task_attribute (pt.top_task_id, 'TASK NAME' ) AS task_name, peia.expenditure_item_date, peia.projfunc_raw_revenue AS func_raw_revenue,
peia.accrued_revenue AS func_accrued_revenue, peia.projfunc_currency_code,
DECODE (peia.project_raw_revenue, NULL,
DECODE (peia.system_linkage_function, 'ST', peia.quantity* NVL (peia.bill_rate, func_get_emp_billrate (pea.incurred_by_person_id, peia.project_id, peia.task_id,
peia.expenditure_item_date )
peia.quantity
peia.project_raw_revenue
( peia.accrued_revenue
/ peia.projfunc_raw_revenue
* peia.project_raw_revenue
) AS project_accrued_revenue,
ppa.project_currency_code,
NVL(peia.bill_rate, func_get_emp_billrate (pea.incurred_by_person_id, peia.project_id, peia.task_id, peia.expenditure_item_date)
) AS bill_rate,
peia.rev_dist_rejection_code, SYSDATE AS generated_date
FROM pa_expenditures_all pea
INNER JOIN (SELECT * FROM per_all_people_f WHERE SYSDATE BETWEEN effective_start_date AND effective_end_date) emp ON NVL(pea.incurred_by_person_id, 61) =emp.person_id
INNER JOIN (SELECT * FROM pa_expenditure_items_all WHERE billable_flag = 'Y' AND quantity <> 0) peia ON pea.expenditure_id =peia.expenditure_id
INNER JOIN (SELECT * FROM pa_projects_all WHERE project_status_code = 'APPROVED' AND distribution_rule IN ('WORK/WORK', 'WORK/EVENT')) ppa ON peia.project_id =ppa.project_id
INNER JOIN pa_tasks pt ON peia.task_id = pt.task_id;
The explain plan for this is:
SELECT STATEMENT CHOOSECost: 26,758 Bytes: 53,373,971 Cardinality: 245,963                          
     9 HASH JOIN Cost: 26,758 Bytes: 53,373,971 Cardinality: 245,963                     
          1 TABLE ACCESS FULL TABLE HR.PER_ALL_PEOPLE_F Cost: 67 Bytes: 546,900 Cardinality: 9,115                
          8 HASH JOIN Cost: 26,581 Bytes: 20,174,029 Cardinality: 128,497                
               2 TABLE ACCESS FULL TABLE PA.PA_TASKS Cost: 596 Bytes: 856,370 Cardinality: 85,637           
               7 HASH JOIN Cost: 25,691 Bytes: 18,889,059 Cardinality: 128,497           
                    3 TABLE ACCESS FULL TABLE PA.PA_EXPENDITURES_ALL Cost: 1,524 Bytes: 5,220,330 Cardinality: 522,033      
                    6 HASH JOIN Cost: 23,701 Bytes: 17,604,089 Cardinality: 128,497      
                         4 TABLE ACCESS FULL TABLE PA.PA_PROJECTS_ALL Cost: 91 Bytes: 49,059 Cardinality: 621
                         5 TABLE ACCESS FULL TABLE PA.PA_EXPENDITURE_ITEMS_ALL Cost: 23,575 Bytes: 68,913,570 Cardinality: 1,188,165
Can anyone suggests what can be done to improve this query?
Thanks in advance
SA
Edited by: user593719 on Oct 5, 2010 10:29 AM
Edited by: user593719 on Oct 5, 2010 10:30 AM

Answer 1

1) Your inner joins are doing record selections that could be done on the Join's ON clause.
2) Look at your functions to see if you can get performance improvements using parameterized CURSORs & adding DETERMINISTIC to the definition.
3) Consider using WITH to first pull your data & then running your functions.
For example:
with
smiths as (
select *
from hr.per_all_people_f papf
where papf.first_name='Smith'
select *
from smiths
where trunc(sysdate) between effective_start_date and effective_end_date;
In general, pare down your data first. Whatever will restrict your data most, start with selecting that table/view.