CONNECT_BY_ROOT with XMLAGG

This is a post about Oracle hierarchical queries. The other day I was after a solution which would display the hierarchy in the execution of some PL/SQL procedures. That is, there was  a main procedure which would call several dozens of other procedures in a predefined order. This order was pre-recorded in a table. What I was after was the ancestral relationship in the order of executions. To demonstrate simply, say there is table called steps, which looks like this:




MAIN_ID                DEPENDANT_ID           
---------------------- ---------------------- 
1                      2
1                      3   
2                      4
2                      3

Process MAIN_ID 1 depends on the execution of 2 and 3 first and then 2 depends on the execution of 4 and 3. Can we execute 1 after we execute 2 and 3?

Well the table says YES!. But how about ancestry? We can clearly see that for process 2 to execute process 4 must execute first! Wouldn't it be nicer, if we could show this ancestry relationship via an SQL query straight out of this table? An SQL query which would show us all dependants of 1 and 2, including parents, granddads etc...

So the results we are after could look like this:

MAIN_ID                DEPENDS_ONTHIS
---------------------- --------------
1                      2,4,3  
2                      3,4 

Here is the query to obtain such hierarchical relationships out of this table by using the CONNECT_BY_ROOT, Oracle 10g operator and some XMLAGG function magic for hierarchical queries:

select main_id, rtrim(xmlagg(xmlelement(d, dependant_id||',')).extract ('//text()'),',') depends_onthis 
from 
(
SELECT  distinct connect_by_root main_id main_id,
    dependant_id
  FROM steps
    CONNECT BY prior dependant_id=main_id
  ORDER BY 1,2
)
group by main_id
order by main_id

MAIN_ID                DEPENDS_ONTHIS
---------------------- --------------
1                      2,4,3  
2                      3,4 

I found this new Operator in Oracle 10g amusing and cool. Here is the Oracle documentation:

And here is the test table scripts if you want to run the above test case for yourself:


drop table steps;

create table steps
(
main_id number,
dependant_id number
);

insert into steps values (1,2);
insert into steps values (1,3);
insert into steps values (2,4);
insert into steps values (2,3);



Materialized View Refresh and the ATOMIC_REFRESH parameter

When refreshing big materialized views in large data warehouses it is always good to check the parameter options available in the DBMS_MVIEW.REFRESH procedure.

Oracle changes the default parameters of its DBMS packages from release to release. I remember back at the times of Oracle 9i a complete refresh would truncate the materialized view, thus the only work that the database was actually doing in a complete refresh, was just an INSERT after the TRUNCATE.

Now in Oracle 10g and Oracle 11g parameters have changed. When there is a COMPLETE materialized view refresh, for the purposes of data preservation, a DELETE is done instead of a TRUNCATE!

The reason for this is because Oracle "changed" the default parameter value of ATOMIC_REFRESH in the DBMS_MVIEW.REFRESH package. In earlier releases the parameter was set to FALSE by default but now it is set to TRUE, which forces a DELETE of the materialized view instead of TRUNCATE, making the materialized view more "available" at refresh time.

But this DELETE is an expensive operation in terms of refresh time it takes. A DELETE is always expensive and sometimes even impossible when we are talking about the complete refresh of materialized views with millions of rows. At COMPLETE refresh a DELETE command takes long time, as it has to make the materialized view available to the users and comply with the ACID theory and its transactional reasons, but who cares? What if we really want to get the refresh done quickly and we don't care about the availability of the materialized view, say in a data warehouse maintenance window?

I have tested how the refresh time improves when you set the ATOMIC_REFRESH => FALSE in the procedure on a 1 CPU home computer with Oracle 11g. If you have more CPUs it will even be quicker because of the PARALLEL parameter.

I just wanted to show the improvement you get in refresh times when ATOMIC_REFRESH is FALSE and Oracle does a TRUNCATE instead of a DELETE at complete refresh. A real time saver with big materialized views in data warehouses. Here is the test:

-- Enable parallel DML

ALTER SESSION FORCE PARALLEL DML


-- Cleanup

-- DROP TABLE BASE_TABLE;




-- Create the Test table

CREATE TABLE BASE_TABLE

(

id NUMBER(10) PRIMARY KEY NOT NULL,

X VARCHAR2(255),

Y VARCHAR2(255)

)

PARTITION BY RANGE

(

ID

)

INTERVAL

(

100000

)

(

PARTITION P1 VALUES LESS THAN (100000),

PARTITION P2 VALUES LESS THAN (200000)

);




-- Make the test table paralllel

ALTER TABLE BASE_TABLE PARALLEL 5

Load the test table with data from a dummy cartesian join so that you get millions of rows. Do you like my loop? Loops only once. Well I had it there to add more millions if I wanted to :-) >

BEGIN

FOR I IN 1 .. 1

LOOP

INSERT /*+ PARALLEL(BASE_TABLE, 5) */ INTO BASE_TABLE SELECT ROWNUM, a.object_name, a.status FROM all_objects a, emp b, dept c;

END LOOP;

COMMIT;

END;




3010392 rows inserted

Create a materialized view log for fast refresh

DROP MATERIALIZED VIEW LOG ON BASE_TABLE;

CREATE MATERIALIZED VIEW LOG ON BASE_TABLE;

Create the fast refresh materialized view

DROP MATERIALIZED VIEW MV_BASE_TABLE;

CREATE materialized VIEW mv_base_table parallel 5 refresh fast

AS

SELECT * FROM BASE_TABLE;

Update the Test table to simulate changing data

UPDATE BASE_TABLE SET Y='INVALID';

COMMIT;

3010392 rows updated

Now you are ready to do the materialized view refresh with the ATOMIC_REFRESH values set to TRUE and then to FALSE. Observe the refresh times. Refresh the materialized view with the two different values in the

ATOMIC_REFRESH parameter.




TRUE case with DELETE

EXEC DBMS_MVIEW.REFRESH(LIST => 'MV_BASE_TABLE', METHOD => 'C', ATOMIC_REFRESH => TRUE);




Elapsed 558.8 seconds

Now is time to do the test with the ATOMIC_REFRESH parameter set to FALSE.




FALSE case with TRUNCATE

Must update again to simulate change in the logs

UPDATE BASE_TABLE SET Y='VALID';

COMMIT;

EXEC DBMS_MVIEW.REFRESH(LIST => 'MV_BASE_TABLE', METHOD => 'C', ATOMIC_REFRESH => FALSE);




Elapsed 215.3 seconds

Half the time! Well if you have real computers it will be much faster. So again we see DELETE is really not good when you work with millions of rows. Nice to have options!

Migrate your Oracle 10g database to ASM

This blog entry is about migrating your Oracle 10g database to ASM using OEM Database Control on ubuntu 9.10 Karmic Koala. To be able to do this you must have:

• CSS service running on your ubuntu box
• ASM instance running
• OEM Database control running

To resolve your problems of starting the Oracle CSS daemon and ASM instance running on Ubuntu Karmic Koala, follow the resources I give below.

You will have to do some hocus-pocus with some parameters to adjust memory, improvise disks with dd (cause I only had one) and CSS script startup hick-ups. Insist with the resources excession.org.uk and pythian.com and oraclebase kindly provide and you will be running ASM on single instance on ubuntu 9.10 Karmic Koala in no time.

You should have the above three ready before you attempt to migrate your databases to ASM. You must go through the trouble and create an ASM instance with DBCA as per the resource below. No ASM instance no migration!

Once you are ready there only 4 steps to follow on the OEM and they are easy!

About ASM

For years DBAs and SAs have been looking over spindles and trying to decide where to put what. Which files shall I put on which disk? RAID this RAID that and so on...

What is the best way to tune I/O for Oracle?

The answer is, easily enough, Let Oracle tune I/O for you with their SAME (Stripe and Mirror Everything) approach. More, the mirroring and striping is done at the extent level and file level and not disk level.

That is exactly what ASM does, it balances I/O activity. It is an Oracle instance which runs alongside your database instances, in memory, and balances the I/O of database data files. It balances and tunes I/O whenever you add disks or you drop them, even online with no intrusion!

All your Oracle datafiles now are controlled by ASM and the way it does it is that ASM makes sure that a file is evenly spread across all disks when the file is allocated, so rebalancing is not required. 


All this sounds very good, if we don't care about I/O tuning what will happen to SQL Tuning? Are we going to do structural and logical tuning only?


Further, with emergence of "Oracle TimesTen In memory" database technology where the whole database can be loaded easily in solid state memory (Flash Disks), I wonder how long will ASM last?


Resources: 


http://www.excession.org.uk/blog/installing-oracle-on-ubuntu-karmic-64-bit.html

http://www.pythian.com/news/810/howto-set-up-oracle-asm-on-ubuntu-gutsy-gibbon/


http://www.oracle-base.com/articles/10g/AutomaticStorageManagement10g.php

Salesforce and force.com

I like salesforce.com and force.com. Both I think are real pioneers in the terra-incognita, or maybe I should call it the nimbus-incognita, which is called the "Cloud-platform". Both are very easy to use SaaS (Software as a Service) platforms. The applications are based on solid data modelling principles and the back-end relational database used is a massive parallel implementation of Oracle databases around the world. You can check their systems availability and the number of transactions done per day here trust.salesforce.com.

Although salesforce insists that their applications are stored in objects which are not "just database tables", their first step in creating an application involves capturing the data model. This is really a very "database" platform.

The platform offers a solid 2x2=4 environment to build and host applications and all work is done with a browser. That is, their IDE is the browser. They even provide you with a sandbox for your unit tests and and Eclipse plug-in for code access to the platform. The IDE is shared between their CRM platform salesforce.com and their developer platform force.com. In the salesforce.com platform you are able to extend a complete and very well designed CRM data model to include anything you wish. In the force.com platform you can create applications from scratch about anything! Among the two, force.com is what impressed me the most and made me to write this blog post. There is not much to be done in the salesforce.com platform, as is a complete working CRM application in itself which would suit any size company which needs a good CRM system.

Force.com on the other hand is a blank canvas. As long as you now what you want, you can start building it straight away. You start with the data model, then you determine your objects, and their relationships and once you have layered your data model in your system, the platform straight away provides you with the means to update and insert records in the back-end database. It automatically builds forms and reports. It also provides you with work-flow and automatic callendaring and emailing functions. Plus, it has a coding language called Apex, which is Java like with classes and triggers enabling you to enhance your application logic. It also has a very HTML like markup language called Visualforce. If you want to migrate your legacy data into the platform you should look at cloud ready ETL and data integration tools like Informatica (£££) and Talend Open Studio  (free).

A project I recently have done in Kizoom using this platform which took me 82 hours that is 10 days approximately to complete. The project was migrating all data and application logic from a legacy database system to force.com. During my work, data integration tools like Informatica and Talend helped me to move the data quickly into the force.com data model, and the already built in features of force.com provided me data entry forms and advanced reports. Because all this is in the cloud I was able to easily utilise geolocation with Google Maps API. The legacy system was a Microsoft SQL Server online database of fault records of computers in publicly available street kiosks around UK. Force.com made the application available in the cloud which meant the UK-wide field engineers of the company were able to easily login and access it on the road without having to use the company network (VPN) but just the cloud.

Below is a snapshot of the app.

Conclusion

The platform is very capable, easy to use and very intuitive. It has tremendous potential and achieves good economics in the development efforts. One shouldn't be shy to use it.

SQL Tuning Advisor and SQL Access Advisor in Oracle 10g

Oracle 10g with its AWR (Automatic Workload Repository) for automatic collection of performance statistics and with ADDM (Automatic Database Diagnostic Monitor) for identifying root causes of problems, is a self-healing and self-tuning getting wiser every day and very soon will start to think on its own database.

There are features in Oracle 10g which help you to quickly find out what is wrong with the database. Features like ASH (Active Session History) to peak into what your active sessions are waiting for and Server Generated Alerts with Metrics and Thresholds are all tools which tell you what is really going on in your database. The OEM (Oracle Enterprise Manager) Database control is a brilliant interface you can use to learn and find out more about all these smart enhancements. All in the name of pro-active database management.

What is database well-being all about anyway? Isn't it high throughput(OLTP) and response time(DSS)? Aren't all databases working to achieve these?

Oracle 10g has also an Advisory Framework which based on workload, will also try to advise you on configuration issues like Buffer Cache, Library Cache, PGA, Undo and Segments used. There are also Tuning-Related Advisors which deal with SQL Tuning and these are:

I - SQL Tuning Advisor
II - SQL Access Advisor

In this post I will try to explain the usage of these two advisors in the job of tuning bad and nasty SQL statements.

All automation is good, self healing databases are very good. But what happens in the development arena when a developer turns up to your desk pointing to an email he/she just sent to you and says '...what's wrong with this SQL statement? it takes ages? Why is Oracle so slow?...' I am particularly fond of the last comment.

How can we quickly look at what is wrong with the SQL statement using the SQL Tuning Advisor? Here is how.

I - SQL Tuning Advisor

Using the DBMS_SQLTUNE package to tune SQL

Assuming the suspect query is something like "SELECT * FROM SESSIONS, SOURCE_EVENTS", two very large tables, and your schema has the ADVISOR privilege granted, you would put this query into the Oracle SQL Tuning Advisor engine using the PL/SQL package DBMS_SQLTUNE like this:

1. Create the tuning task

In SQL*Plus and in the schema where objects live, use the procedure dbms_sqltune.create_tuning_task to create a tuning task.

SQL>
declare tname varchar2(30);
begin
tname := dbms_sqltune.create_tuning_task(sql_text=>'select * from sessions, source_events');
end;
/

PL/SQL procedure successfully completed.

Then you can check to make sure that your task is created and get the TASK_NAME value as you will need this in the next step. I could name the task, just being lazy...

SQL> select task_name from user_advisor_log;

TASK_NAME
------------------------------
TASK_51415


2. Execute the tuning tasks

Execute the tuning task you created in step 1 with the procedure dbms_sqltune.execute_tuning_task

SQL> exec dbms_sqltune.execute_tuning_task('TASK_51415');



3. Get the tuning report

Get the tuning report/recommendation with the following SQL.

SQL> set long 50000
SQL> select dbms_sqltune.report_tuning_task('TASK_51415') from dual;


DBMS_SQLTUNE.REPORT_TUNING_TASK('TASK_51415')

(CLOB) GENERAL INFORMATION SECTION
-------------------------------------------------------------------------------
Tuning Task Name : TASK_51415
Tuning Task Owner : ACHILLES
Scope : COMPREHENSIVE
Time Limit(seconds) : 1800
Completion Status : COMPLETED
Started at : 12/20/2009 18:44:05
Completed at : 12/20/2009 18:44:14
Number of SQL Restructure Findings: 1

-------------------------------------------------------------------------------
Schema Name: ACHILLES
SQL ID : 8qsdf0a6qv4g7
SQL Text : select * from sessions, source_events

-------------------------------------------------------------------------------
FINDINGS SECTION (1 finding)
-------------------------------------------------------------------------------

1- Restructure SQL finding (see plan 1 in explain plans section)
----------------------------------------------------------------
An expensive cartesian product operation was found at line ID 3 of the
execution plan.

Recommendation
--------------
- Consider removing the disconnected table or view from this statement or
add a join condition which refers to it.

Rationale
---------
A cartesian product should be avoided whenever possible because it is an
expensive operation and might produce a large amount of data.

-------------------------------------------------------------------------------
EXPLAIN PLANS SECTION
-------------------------------------------------------------------------------

1- Original
-----------
Plan hash value: 3883039950

---------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop | TQ |IN-OUT| PQ Distrib |
---------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 33T| 4158T| 4367M (2)|999:59:59 | | | | | |
| 1 | PX COORDINATOR | | | | | | | | | | |
| 2 | PX SEND QC (RANDOM) | :TQ10001 | 33T| 4158T| 4367M (2)|999:59:59 | | | Q1,01 | P->S | QC (RAND) |
| 3 | MERGE JOIN CARTESIAN | | 33T| 4158T| 4367M (2)|999:59:59 | | | Q1,01 | PCWP | |
| 4 | SORT JOIN | | | | | | | | Q1,01 | PCWP | |
| 5 | PX RECEIVE | | 8439K| 829M| 3175 (2)| 00:00:39 | | | Q1,01 | PCWP | |
| 6 | PX SEND BROADCAST | :TQ10000 | 8439K| 829M| 3175 (2)| 00:00:39 | | | Q1,00 | P->P | BROADCAST |
| 7 | PX BLOCK ITERATOR | | 8439K| 829M| 3175 (2)| 00:00:39 | 1 | 16 | Q1,00 | PCWC | |
| 8 | TABLE ACCESS FULL| SESSIONS | 8439K| 829M| 3175 (2)| 00:00:39 | 1 | 16 | Q1,00 | PCWP | |
| 9 | BUFFER SORT | | 4013K| 122M| 4367M (2)|999:59:59 | | | Q1,01 | PCWP | |
| 10 | PX BLOCK ITERATOR | | 4013K| 122M| 532 (4)| 00:00:07 | 1 | 16 | Q1,01 | PCWC | |
| 11 | TABLE ACCESS FULL | SOURCE_EVENTS | 4013K| 122M| 532 (4)| 00:00:07 | 1 | 16 | Q1,01 | PCWP | |
---------------------------------------------------------------------------------------------------------------------------------------

1 rows selected


Well done SQL Tuning Advisor, a good answer. It immediately spotted the Cartesian product and it recommends with a rationale that it would be expensive!


II - SQL Access Advisor

SQL Access advisor is another Advisory Framework tool which provides primarily advice on the creation of indexes, materialized views, materialized view logs to improve query performance and response time.

Both SQL Tuning Advisor and SQL Access Advisor tools are quite powerful as they can source the SQL they will tune automatcially from multiple different sources, including SQL cache, AWR, SQL tuning Sets and user defined workloads. Look at the Oracle documentation to find out how all this is possible under OEM Database Control.

What I focus on here is how to quickly pass through these tools ad-hoc SQL that the developer will bring to you and will demand to be tuned there and then.

So using the QUICK_TUNE procedure from the DBMS_ADVISOR package here is how you would tune a nasty single SQL statement for better data access paths with SQL Access Advisor.

1. Create tuning task

In the schema where the table is run the following

VARIABLE tname VARCHAR2(30);
VARIABLE sql_stmt VARCHAR2(30);
EXECUTE :sql_stmt := 'select count(*) from sessions where short_code=''abracadambra'' ';
EXECUTE :tname := 'myquicktunetask';
EXECUTE DBMS_ADVISOR.quick_tune(dbms_advisor.sqlaccess_advisor, :tname, :sql_stmt);


2. Check to see the task is there

Again to see the task myquicktunetask which you just created run in the schema

SQL> select task_name from user_advisor_log
TASK_NAME
------------------------------
TASK_51415
myquicktunetask

2 rows selected


3. See tuning advice

Now, to see the advice you get you will have to look in the DBA dictionary view
DBA_ADVISOR_ACTIONS. So login as DBA and filter for the task_name='myquicktunetask' in the DBA_ADVISOR_ACTIONS dictionary view. Well, you can view the advice in the USER_ADVISOR_ACTIONS dictionary view as well.

SELECT command,
attr2 ,
attr3 ,
attr4 ,
attr5
FROM dba_advisor_actions
WHERE TASK_NAME='myquicktunetask'

TASK_NAME||'|'||COMMAND||'|'||ATTR2||'|'||ATTR3||'|'||ATTR4||'|'||ATTR5
(CLOB) myquicktunetask|CREATE MATERIALIZED VIEW LOG||ROWID, SEQUENCE| INCLUDING NEW VALUES|("SHORT_CODE")
(CLOB) myquicktunetask|CREATE MATERIALIZED VIEW||REFRESH FAST WITH ROWID|ENABLE QUERY REWRITE|SELECT ACHILLES.SESSIONS.SHORT_CODE C1, COUNT(*) M1 FROM ACHILLES.SESSIONS WHERE (ACHILLES.SESSIONS.SHORT_CODE = 'tl') GROUP BY ACHILLES.SESSIONS.SHORT_CODE
(CLOB) myquicktunetask|GATHER TABLE STATISTICS||-1||


Conclusion

Well I find both tools extremely useful. Both are good starting points in resolving SQL tuning issues and discussions. SQL Tuning Advisor seems to be doing logical optimization mainly by checking your SQL structure and statistics and the SQL Access Advisor does suggest good data access paths, that is mainly work which can be done better on disk.