Oracle SQL performance with database links - db link

What are some of the performance issues when running SQL across a database link (dblink)?

  • Pull vs. Push - In general, performance can be faster if you "pull" the data (calling the remote table from the master instance), as opposed to a "push" where you perform the join on the remote table.  This is especially true of you have a large sort, because the rows may be transferred to the remote host for sorting, and then back again afterwards.
  • driving_site hint - The driving site hint forces query execution to be done at a different site than the initiating instance.  This is done when the remote table is much larger than the local table and you want the work (join, sorting) done remotely to save the back-and-forth network traffic.  n this example, we use thedriving_site hint to force the "work" to be done on the site where the huge table resides:
select /*+DRIVING_SITE(h)*/ 
  tiny_table        t, 
  huge_table@remote h
   t.deptno = h.deptno;

Note:  The driving_site hint will not work with CTAS (create table as select) and with create materialized view syntax because they are DDL and these operations must take place on the original instance.

  • Views - Some recommend creating a view on the remote site referencing the local tables and calling the remote table via the local view. 
create view local_cust as select * from cust@remote;

You can get the same effect by using an inline view:

FROM (SELECT stuff FROM emp@remote) a
  • Sorting - If your SQL performs a sort, be aware that the sort will be performed on the LOCAL database.  This is one reason why it's bad to use a "push" approach, because the rows will traverse back-and-forth. 

  • Parallelism - Parallel query across a database link can be quite complex.   In a distributed environment, pieces of a table may reside on many remote servers.  For example, assume that we have a distributed architecture where local customer tables are kept on each instance.  You could access all of the remote rows in a single query, using inter-instance parallel execution.  In this example, the query is executed from the north_carolina instance, accessing two remote instances in-parallel:
select customer_name, sum(purchase_amount) from sales
   select customer_name, sum(purchase_amount) from sales@san_francisco
   select customer_name, sum(purchase_amount) from sales@new_york
   group by
  • In this case the north_carolina instance drives the distributed parallel query and it is the north_carolina instance that must gather and sort the result set.

Alberto Dell'Era notes this tip for distributed query parallelism:

"I had some problems myself while mixing db-links and parallel operations; i solved them perfectly by creating a view at the remote site with the same text as the query, and then performing a "select *" from the view. "

Also, note that there are many hidden parameters that influence OPQ performance (always consult Oracle technical support before changing an undocumented parameter)

NAME                                VALUE
----------------------------------- ---------------------_parallel_adaptive_max_users        1
_parallel_default_max_instances     1
_parallel_execution_message_align   FALSE
_parallel_fake_class_pct            0
_parallel_load_bal_unit             0
_parallel_load_balancing            TRUE
_parallel_min_message_pool          64560
_parallel_recovery_stopat           32767
_parallel_server_idle_time          5
_parallel_server_sleep_time         10
_parallel_txn_global                FALSE
_parallelism_cost_fudge_factor      350

Monitoring parallel query for remote distributed queries is also challenging.

  • Network -  Generally, the limiting factor in SQL performance of a distributed query (over a db-link) is the speed of the network (make sure that you have a high-speed network with proper SDU), but there are other issues:
    • tnsnames.ora - Some recommend a separate listener and a larger value for SDU to accommodate jumbo Ethernet frames.

"By using a separate listener for the gigabit network with a larger SDU - you can better exploit jumbo Ethernet frames. beware - interoperability just plain sucks.

Credit :: Tips by Burleson Consulting)

Oracle Coding Best Practices

This document defines a list of standards and guidelines that will be used when submitting PL/SQL Components. Good standards do not constrain the creativity of designers and developers, but rather encourage the development of best practices.  

Read More

Speed Oracle SQL with Temporary Tables (GTT)

The effective use of temporary tables while doing huge batch process can boost performance. The concept is to store intermediate data while performing huge processing in temporary tables. Temporary table store data in-memory that results in better and stable SQL plan for complex summarization activities.

#GTT #SQL Performance #Oracle

Read More

What is the key to Oracle SQL Performance ?

Oracle SQL performance tuning is multidimensional in nature since it involves a lot of variables like underlying operating system, Oracle Server, network and SQL.

We are going to discuss SQL performance( assuming the other variables are working fine) and look at two key points that can help debug a bad SQL. To arrive at the two key points, we need to look how database technology has evolved lately.

SQL Performance

Read More

1000's of table

Came across this article on managing relational data from a very different perspective. Typically relational design in OLTP grows to 100’s of tables. When we have 100’s of table, we require each business change to have schema changes defined upfront, are more difficult to use with agile methods and a challenge to use in a continuously integrated environment without significant additional engineering.

SQL Performance  Design

Read More

SQL Profiles Vs SQL Plan Baselines?

What's difference between SQL profiles and SQL plan baselines. But first let's briefly recap each feature.

Programmers add hints to SQL query to give direction to the Optimizer. Both SQL Profiles and Baselines are nothing but collections of stored hints that “attach” to their target SQLs.

Both SQL Profiles and SPM Baselines are designed to deal with the same problem "Optimizer sometimes produces a very inefficient execution plan", and they are both doing it by essentially abandoning the idea that “all SQLs are created equal”. Instead, another idea is put forward  “Some SQLs are special and deserve individual treatment”.

The way they are implemented, both SQL Profiles and SPM Baselines are:

  • External objects that contain additional “magic” information for the optimizer
  • Attached to individual SQLs and influence them only
  • Built by actually running the SQL and using feedback from runtime engine

What is the “magic” contents that profiles and baselines use to influence SQL execution plans ?

Based on hint contents, there are some major differences in SQL influence “mechanics” between SQL Profiles and SPM Baselines:

  • SQL Profiles (“Soft power”) are mostly ADVISORS that work with the Optimizer making its estimations more precise
  • SPM Baselines (“Hard Power”) on the other hand are ENFORCERS. They completely ignore the Optimizer and simply tell the runtime engine what to do

Is there any difference in purpose ?

Now that we’ve seen the actual contents of both SQL Profiles and SPM Baselines, let’s talk about their purpose.

SQL Profiles were designed to correct Optimizer behavior when underlying data do not fit anymore into its statistical (and simplistic) world view. Their goal is to create the absolute best execution plan for the SQL by giving the very precise data to the optimizer. In other words, SQL Profiles are all about “ON THE SPOT” SQL PERFORMANCE.

Baselines, on the other hand are different. They were designed to prevent execution plans from changing uncontrollably and their goal is to only allow execution plans that were proven to be efficient. In other words, SPM Baselines are all about LONG TERM STABILITY.

Baselines know what plan they are trying recreate and SQL Profiles do not. SQL Profiles will blindly apply any hints it has and what you get is what you get. Baselines will apply the hints and if the optimizer gets the plan it was expecting, it uses the plan. If it doesn't come up with the expected plan, the hints are thrown away and the optimizer tries again (possibly with the hints from another accepted Baseline).

Profiles have a “force matching” capability that allows them to be applied to multiple statements that differ only in the values of literals. Think of it as a just in time cursor sharing feature. Baselines do not have this ability to act on multiple statements.

Can Baseline be disregarded?

Baselines can be rejected in the underlying objects (“indexes”) were missing. Baselines can be disregarded, even without such extreme conditions as a specified index having been removed. The reason for this is that Baselines attempt to apply enough hints to limit the choices the optimizer has to a single plan, but there are situations where the set of hints is not sufficient to actually force the desired plan. What I mean is that the hints will eliminate virtually all possibility, but there still may be a few that are valid and so it’s possible to get a different plan. In fact, there are situations where the act of creating a Baseline causes the plan to change. This is clearly not intended behavior, but it can happen.  At any rate, in these cases where the optimizer arrives at a different plan than was expected, the optimizer uses a plan generated completely without the hints (or possible picks another accepted Baseline and applies it’s hints). This behavior is very different from SQL Profiles which blindly apply the hints.

Note: SQL Profiles can also be created manually with the same set of hints that would be used in a Baseline and there is a MOS note regarding how to do this. 

SQL query to get SPM baseline.

SQL> select count(*) from dba_sql_plan_baselines where parsing_schema_name='$1';




SQL Tuning Hierarchy - Visual

A picture is worth a thousand words.
— Napolean Bonaparte