Question related to table partitioning

Question

Dear Friends,
I have two tables:
Table 1.
SQL> desc ProcessedData_Details
Name Null? Type
VERSIONNO NOT NULL FLOAT(126)
VARCHAR_VALUE VARCHAR2(255)
NUMERIC_VALUE FLOAT(126)
DATE_VALUE DATE
BOOLEAN_VALUE VARCHAR2(10)
REPORTINGDATE_ID NOT NULL NUMBER
SYSTEMDEAL_ID NUMBER
OUTPUT_TEMPLATE_ID NUMBER
SECTION_ID NUMBER
DATAPOINT_ID NUMBER
PROCESSDATA_ID NOT NULL NUMBER
ROW_ID NUMBER
USER_ID NUMBER
ROLE_ID NUMBER
IS_STORE NOT NULL NUMBER
Table 2.
SQL> desc Reporting_Date
Name Null? Type
REPORTINGDATE_ID NOT NULL NUMBER(10)
REPORTINGDATE DATE
CURRENTPERIOD NUMBER(38)
Both tables are joined using 'REPORTINGDATE_ID'. Now I've to partition Table 1 based on the column 'REPORTINGDATE' of Table 2. How can it be done?
Edited by: Santosh Kumar on Dec 12, 2008 2:35 PM

Answer 1

updates to the status column is frequent, considering data inflow.
status has dfferent values like new,active,inactive etc.
for the entity record, workflow, has different states.
table can be partitioned on status field,
for a new state of a workflow record, there will be a new row and the old record is updated with a different status, hence row movement across paritions
for eg:
workflow_id=100 (internal id
workflow_parent=WF_XXXX01
status=new
next time, if there is a status change for the workflow, WF_XXXX01
workflow_id=101
workflow_parent=WF_XXXX01
status=active
etc
if PK to be a local index, then it will be workflow_id+status
possible query patterns,
select * from workflow where workflow_id=id_val and status='new'
select * from workflow where <some_other_criteria>
select * from workflow where status='new'
select * from workflow where status='active'
the only advantage i see here is to keep separate partitions for each status to group records, for eg. instead of keeping a separate entity to keep the inactive records.
if the predicate includes the status field, it needs to touch only the corresponding partition, if local index is there, then there can be performance benefits.
but global index will be ok, if there are limited number of partitions and query retrieves only small number of rows and response time is critical
may be after some time, as part of puring, the inactive paritions can be dropped, if there are only local indexes, then there is no parition maintenance overhead
but considering overall overhead, in terms of row contention related issues,
for an OLTP, would you suggest partitioning?
thanks,
charles
Edited by: user570138 on 2-nov-2010 8:10
Edited by: user570138 on 2-nov-2010 8:18
Edited by: user570138 on 2-nov-2010 8:18