An Introduction To Oracle SQL Tuning

Chris Adkin 30th May 2008

01/05/2012

1

Some Inspirational Thoughts Before We Begin . . .

The following Ask Tom excerpt comes in response to a “You Asked Can u give a methodology of tuning the sql statements.” question. The link to the full answer is at: http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_I D:8764517459743 Despite being five years old, in the intervening time, artificial intelligence has not been built into the RDBMS where by it can re-write your SQL such that it will run in the most efficient manner possible. Advisors take some of the leg work out of tuning and the tools such as DBMS_XPLAN, v$ views etc constantly change, evolve and improve, however SQL tuning and writing efficient SQL is not a prescriptive process that can be captured in a process. I will however try to present useful techniques and good practise to „demystify‟ some of the „art‟ behind this.
2

Some Inspirational Thoughts Before We Begin . . .

1.1 Efficient SQL This was probably the hardest part of the book to write - this chapter. That is not because the material is all that complex, rather because I know what people want - and I know what can be delivered. What people want: The 10 step process by which you can tune any query. What can be delivered: Knowledge about how queries are processed, knowledge you can use and apply day to day as you develop them. Think about it for a moment.
3

Oracle Enterprise Manager with its tuning pack. suggest materialized views.  If there were a 10 step or even 1. 4 .Some Inspirational Thoughts Before We Begin . Oh don't get me wrong. SQL Navigator and others. . offer to add hints to the query to try other access plans. we would write a program to do it. What they do is primarily recommend indexing schemes to tune a query.000 step process by which any query can be tuned (or even X% of queries for that matter). . there are many programs that actually try to do this .000.

In fact.which if they were universally applicable . They offer "rules of thumb" (what I generally call ROT since the acronym and the word is maps to are so appropriate for each other) SQL optimizations .Some Inspirational Thoughts Before We Begin .the optimizer would do it as a matter of fact.it rewrites our queries all of the time. These tuning tools use a very limited set of rules that sometimes can suggest that index or set of indexes you really should have thought of during your design. the cost based optimizer does that already . .  They show you different query plans for the same statement and allow you to pick one. . 5 .

understand the question being asked and process the question .if there were an N step process to tuning a query. .Some Inspirational Thoughts Before We Begin .the optimizer would incorporate it all and we would not be having a discussion about this topic at all.  I'll close this idea out with this thought . to writing efficient SQL .rather then syntax. 6 . It is like the search for the holy grail maybe someday the software will be sophisticated enough to be perfect in this regards. it will be able to take our SQL. .

Section 1 Mechanics Of The Cost Based Optimizer 01/05/2012 7 .

Devises the best plan for executing a query based on cost.What Is The Optimizer      I will focus on the Cost Based Optimizer. except when the „wrong‟ plan is selected. Around since Oracle 7. The stages of optimisation will be covered on the next set of slides. Transparent to application and users. 8 .

these are affectively different plans for the same SQL text when different variables are supplied (an over simplification. prior releases use a combination of SQL address and hash value. but basically what happens).Stages Of Optimisation     On 10g parsed SQL statements are assigned unique identifiers called sql ids. If a query does not exist in parsed form in the shared pool. Oracle can also perform optimisations at run time. but basically what happens). 9 . a hard parse or optimisation takes place (a gross over simplification. A parsed SQL statement (cursor) can have multiple child cursors.

Stages Of Optimisation  Establish the environment Ascertain what parameters are set  Ascertain what bug fixes are in place based on the setting of the compatible parameter   Query transformation Sub query un-nesting  Complex view merging  Set Join Conversion  Predicate move around  10 .

Join order consideration.Stages Of Optimisation    Establish the base statistics of all the relevant tables and indexes. 11 . Single table access cardinality estimation.  Record will be kept of the best join order maintained so far.

What Influences The Cost Based Optimizer ?    Object statistics Object data types Oracle initialisation parameters. refer to:v$sys_optimizer_env  v$ses_optimizer_env  V$sql_optimizer_env  12 .

What Influences The Cost Based Optimizer ?  System statistics:COMPLETED 11-26-2006 19:53 11-26-2006 19:53 1 1081.aux_stats$ SQL> / SYSSTATS_INFO STATUS SYSSTATS_INFO DSTART SYSSTATS_INFO DSTOP SYSSTATS_INFO FLAGS SYSSTATS_MAIN CPUSPEEDNW SYSSTATS_MAIN IOSEEKTIM SYSSTATS_MAIN IOTFRSPEED SYSSTATS_MAIN SREADTIM SYSSTATS_MAIN MREADTIM SYSSTATS_MAIN CPUSPEED SYSSTATS_MAIN MBRC SYSSTATS_MAIN MAXTHR SYSSTATS_MAIN SLAVETHR 13 .76881 10 4096 1* select * from sys.

set this to IO or CPU.What Influences The Cost Based Optimizer ?       System statistics facilitate something called the “CPU costing model” introduced in Oracle 9i. In Oracle 9i. Stats that are not populated in aux_stats$ only appear after system statistics have been gathered. 14 . Until this came along the optimizer did not take into account CPU performance nor the difference in performance between single and multi block reads. In 10g “out of the box” statistics called “no workload” statistics are provided. no system statistics are present. The optimizer cost model can be set via _optim_cost_model.

oracle.What Influences The Cost Based Optimizer ?  !!! Warning !!! You have to be very careful when gathering system statistics.  This can seriously skew a plan‟s costings. Oracle supply an “I/O calibration” tool called Orion which might help here: http://www.  If you have any I/O sub system with a cache you can end up getting single block and multi block I/O times that reflect the speed of the cache rather than the under lying disks.html   15 .com/technology/software/tech/orion/index.

What Influences The Cost Based Optimizer  SQL Usage Unless the NO_MONITOR hint is used.  Oracle can get confused with what to do for columns where the LIKE and col_usage$ and hence not create histograms when size auto is specified. col_usage$ in the data dictionary will be updated whenever a statement is parsed.  16 .  DBMS_STATS uses this in order to determine whether a histogram should be created on a column when size = AUTO is specified in the method_opt.

partitioning. . constraints etc . .What Influences The Cost Based Optimizer  Hints and profiles SQL hints  Stored out lines (essentially hints)  SQL profiles as created by the Oracle 10g tuning advisor   Objects. 17 . presence of indexes.

Section 2 A Plan Of Attack For Investigating A Bad Plan 01/05/2012 18 .

Identify The SQL To Be Tuned  Many ways of doing this:Automatic Database Diagnostic Manager $ORACLE_HOME/rdbms/admin/addmrpt  Automatic Workload Repository reports $ORACLE_HOME/rdbms/admin/awrrpt  Toad  SQL Tuning Advisor $ORACLE_HOME/rdbms/admin/sqltrpt  19 .

for previous releases this a combination of hash value and sql address. user_io_wait_time. As a first pass tuning try running the SQL tuning advisor on the query. my preferred way of doing this is via the sqltrpt script in $ORACLE_HOME/rdbms/admin. cpu_time.A Simple Script For Identifyng SQL To Be Tuned  SELECT * FROM (SELECT sql_id. elapsed_time.   20 . concurrency_wait_time FROM v$sql ORDER BY elapsed_time DESC) WHERE rownum <= 10 In 10 onwards SQL statements are uniquely identified by sql ids.

Optimizer flaws. 21 . more on this later.2.4.0. SQL that is “optimizer unfriendly”.0. more on this later. such as the predicate independence assumption. there are the best part of 100 bug fixes related to the optimizer in 10. Lack of appropriate indexes to support your query Bugs in the optimizer. SQL that is inefficient by design.Causes Of Poor Performance       Stale or missing statistics (includes missing histograms).

22 . Abuse of the Oracle optimizer environment. Poor schema design.Causes Of Poor Performance     Index. Misuse or poor use of Oracle features. table type or partitioning schemes that are not conducive to good performance with the query mix you are running.

 Is the part of the application running slowly because it is not designed to deal with the shape and / or size of the data it is processing ?.The Tuning Ethos  Ask the following questions:Has what I need to tune ever ran „fast‟.  Do not confuse something that is running slow because it has an abnormally high workload for something that is inefficient at what it does.  23 . if so what has changed in the environment.

if you have a problem with one query. for accurate and consistent response times avoid database and servers which are being used by other users if possible.g. Use realistic test data. Try to use a clean environment without factors that may skew your results. a global change such as a change to an initialisation parameter is unlikely to address the root cause of the issue and it may cause performance regression for queries that are currently performing well.g. Try to make the scope of any changes match the scope of the problem. e. 24 .The Tuning Ethos  When tuning     Only make one change at a time so as to be able to accurately gauge what affects your query. e.

 Object types being changed. e.  SQL being re-written. heap to index organized tables.  The use of hints.  Schema design changing.  Statistic gathering regimes being modified.g.g. e. extra columns added to tables in order to reduce numbers of joins.  25 .The Tuning Ethos  Tuning a piece of SQL may result in:Indexes being added or modified.

g.  26 . e. . col NOT IN (1. .  Recommends SQL profiles. .Tuning Is Dead We Have The Tuning Advisor !!!  Based on my experience this:Recommends indexes  Identifies stale statistics  Notifies you of Cartesian joins  Spots predicates not conducive to the efficient use of indexes. col <>.  Identifies parts of statements where view mergining cannot take place. 2.

Tuning Is Dead We Have The Tuning Advisor !!!      If the tuning advisor does anything else. It is the best tool of its type that Oracle have produced to date. I am yet to see it. it is not artificially intelligent and it smacks of the „ROT‟ tools that Tom Kyte mentioned. SQL profiles are enhanced hints that provide the optimizer with extra information with which to get cardinality estimates correct with. The use of SQL profiles can cause a problem if your base data changes such that they are no longer relevant. 27 . This is good for a first pass at identifying more obvious causes of a query running sub optimally.

Some people instantly assume that because a plan contains a full table scan that the execution plan is therefore bad.A Simple „Methodology‟ For Query Tuning     Despite what the Oracle marketing machine may say. This is an out dated method of tuning with origins based around the rule based optimizer. The tuning advisor in its 10g incarnation is no substitute for a knowledgeable DBA at best it semi automates the task of tuning. 28 . tuning is not a prescriptive process.

The table is „small‟ in size‟. result in all or a large proportion of a tables data being returned.e. Reasons for full table scans might include:    Queries that use un-selective predicates. The indexes you think the query should use have poor clustering factors. 29 . I. Queries return most of the columns in a table.Good And Bad Plans   Do not automatically jump to the conclusion that a bad plan contains full table scans and a good plan only uses index range scans.

30 . you will not always be asked for the same information or to use the same diagnostic tools depending on who the service request is assigned to. There is no such thing as a definitive tuning methodology as alluded to by Tom Kyte. This may change with the DBMS_SQLDIAG package in 11g. Even when Oracle support is engaged with a tuning related service request. aka the SQL test case builder.A Simple „Methodology‟ For Query Tuning     What I will outline is a plan of attack for tackling query issues.

. 31 . partitioning schemes etc . Oracle has produced sub optimal plans due to discrepancies between predicated cardinalities and actual cardinalities. .A Simple „Methodology‟ For Query Tuning  My “plan of attack” will not cover the writing of efficient SQL. this would require a different presentation entirely. are in place. I will assume:    All reasonable indexes. Nothing daft has been done to abuse the optimizer environment.

A Simple „Methodology‟ For Query Tuning 1.display)  A better way of explaining the execution plan is by taking it straight from the shared pool:SQL> SELECT * FROM table(dbms_xplan. I. Obtain the execution plan for your query  If you run an explain plan from the SQL prompt this will give you the plan that Oracle predicts the statement will use.e.display_cursor(‘sqlid’)) Sql ids are new to Oracle 10g and uniquely identify parsed SQL statements 32 .:SQL> EXPLAIN PLAN FOR SELECT * FROM dual SQL> SELECT * FROM table(dbms_xplan.

33 . NULL.DISPLAY_AWR(<your SQL id goes here>)  If you use explain plan. this will give you what the optimizer predicts the plan will be. Obtain the execution plan for your query   The DBMS_XPLAN package is the best tool here.A Simple „Methodology‟ For Query Tuning 1. this may not necessarily but what the plan is when the query runs due to such things as bind variable peeking. ‘ALLSTATS LAST’)  Obtain the plan from the shared pool SELECT * FROM TABLE(DBMS_XPLAN.DISPLAY_CURSOR(NULL.DISPLAY_CURSOR(<your SQL id goes here>)  Obtain the plan from the workload repository SELECT * FROM TABLE(DBMS_XPLAN. Where possible obtain the plan after running the query: SELECT * FROM TABLE(DBMS_XPLAN.

"DEPTNO"="DEPT"."SAL"<="SALGRADE".filter("EMP"."SAL">="SALGRADE".Understanding A Basic Execution Plan PLAN_TABLE_OUTPUT ----------------------------------------------------------------| Id | Operation |Name |Rows|Bytes|Cost | ----------------------------------------------------------------| 0 | SELECT STATEMENT | | | | | | 1 | NESTED LOOPS | | | | | | 2 | NESTED LOOPS | | | | | | 3 | TABLE ACCESS FULL | SALGRADE| | | | |* 4 | TABLE ACCESS FULL | EMP | | | | | 5 | TABLE ACCESS BY INDEX ROWID| DEPT | | | | |* 6 | INDEX UNIQUE SCAN | DEPT_PK | | | | ----------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------- 4 ."DEPTNO") 34 ."HISAL" AND "EMP"."LOSAL") 6 .access("EMP".

35 . . the table for which the filter predicates will reduce the most.A Simple „Methodology‟ For Query Tuning 2. I. this is the first table in the join order and usually the table that is furthest indented into the plan. Is the most restricted table the „driving‟ table. not always an issue.e. Is the access for the driving table via a full table scan. but could indicate that the join order is incorrect Continued . Visually sanity check your execution plan    Does the plan contain any Cartesian joins ?. .

ID_JOB FROM TR_JOB A.A Simple „Methodology‟ For Query Tuning 2.SD_JOB_STATUS = :2 AND B. .'yyyy-MM-dd hh24:MI:SS') AND F.'yyyy-MM-ddhh24:MI:SS') AND C.APPMT_START_TIME < to_date(:4.ID_JOB=A. TR_CONSUMPTION_HISTORY B WHERE A. TR_OFFCYCLE_JOB_DETAILS F. TR_CYCLIC_JOB_DETAILS C.VISIT_DATE_FROM < to_date(:6. Visually sanity check your execution plan  Are you battling against the schema design.ID_JOB=A.ID_JOB )) ---------------------------------------------------------------------------------------------| Id | Operation | Name | Rows | Bytes | Cost | ---------------------------------------------------------------------------------------------| 0 | SELECT STATEMENT | | 1 | 151 | 1187 | | 1 | HASH UNIQUE | | 1 | 151 | 1187 | | 2 | NESTED LOOPS | | 1 | 151 | 1176 | | 3 | MERGE JOIN CARTESIAN | | 1 | 104 | 684 | | 4 | NESTED LOOPS | | 1 | 51 | 665 | |* 5 | INDEX FAST FULL SCAN | TR_CONSUMPTION_HISTORY_IX2 | 1 | 5 | 663 | |* 6 | TABLE ACCESS BY INDEX ROWID| TR_JOB | 1 | 46 | 2 | |* 7 | INDEX UNIQUE SCAN | TR_JOB_ID_PK | 1 | | 1 | | 8 | BUFFER SORT | | 32474 | 1680K| 682 | | 9 | TABLE ACCESS FULL | TR_OFFCYCLE_JOB_DETAILS | 32474 | 1680K| 19 | |* 10 | TABLE ACCESS FULL | TR_CYCLIC_JOB_DETAILS | 1 | 47 | 492 | ---------------------------------------------------------------------------------------------- Continued . 36 .APPMT_END_TIME > to_date(:5.ID_JOB AND A. an example:SELECT DISTINCT A.ID_JOB ) OR ( C.'yyyy-MM-dd hh24:MI:SS') AND F.ID_SERVICEPOINT =:1 AND A.'yyyy-MM-dd hh24:MI:SS') AND C.SD_READING_TYPE = :3 AND (( F.VISIT_DATE_TO > to_date(:7. .ID_JOB = B.

specifically if one of the branches evaluates to TRUE the join predicates do not get performed for two tables. We have an index fast full scan (index equivalent of a full table scan) on TR_CONSUMPTION_HISTORY_IX2. note that the only reason we are going to the TR_CONSUMPTION_HISTORY table is to get the contents of the sd_reading_type column.A Simple „Methodology‟ For Query Tuning 2. Visually sanity check your execution plan In the previous example there are two issues:  There is a Cartesian join due to the OR condition. 37 .

Check the ‘environment’ is correct for the query:  Are statistics up to date for all the objects used by the query. you can run the sqltrpt script from $ORACLE_HOME/rdbms/admin. this will also highlight any tables with stale statistics. supply the sql id for your query and the tuning advisor will go to work for you.A Simple „Methodology‟ For Query Tuning 3. Do appropriate indexes exists. 38 . Continued . . .

A Simple „Methodology‟ For Query Tuning
3. Checking the ‘environment’ is correct for the query From Greg Rahn of the Oracle Real World performance group:I think it’s important to understand what variables influence the Optimizer in order to focus the debugging effort. There are quite a number of variables, but frequently the cause of the problem ones are: (1) non default optimizer parameters and (2) nonRepresentative object/system statistics. Based on my observations I would say that the most abused Optimizer parameters are:
  

OPTIMIZER_INDEX_CACHING OPTIMIZER_INDEX_COST_ADJ DB_FILE_MULTIBLOCK_READ_COUNT
39

A Simple „Methodology‟ For Query Tuning
3. Checking the ‘environment’ is correct for the query
Many see setting these as a solution to get the Optimizer to choose an index plan over a table scan plan,but this is problematic in several ways:  This is a global change to a local problem Although it appears to solve one problem, it is unknown how many bad execution plans resulted from this change.  The root cause of why the index plan was not chosen is unknown, just that tweaking parameters gave the desired result  Using non-default parameters makes it almost impossible to correctly and effectively troubleshoot the root cause

40

A Simple „Methodology‟ For Query Tuning
4. Understand your data, how ‘Selective’ is it ?

Are the predicates and variables used in your statement going to retrieve a „significant‟ or „small‟ proportion of rows from the queries base tables ?. The following predicates come from a query I have investigated the performance of:UPPER (a.walkroute_reviewed_flag) = 'Y' ( ( a.sd_service_type = '10000' AND a.sd_servicepoint_status = '10001' ) OR ( a.sd_service_type = '10001' AND a.sd_occupied_status = '10001' AND a.sd_servicepoint_status = '10001' ) OR a.sd_servicepoint_status = '10000' OR a.sd_servicepoint_status = '10003‘

It returns a third of the data in the AD_SERVICE_POINT table, therefore if the clustering factors of the appropriate indexes are not particularly good, a full table scan is probably the most efficient way to retrieve the relevant data.41

 This method of trouble shooting is sometimes referred to as “Tuning by cardinality feedback”. will be out by an order of magnitude and „ripple‟ throughout the rest of the place. 42 . Are the predicted cardinalities for for the execution plan accurate  If the predicted cardinalities for the row sources in the plan are accurate and the appropriate access paths (indexes) exist. detailed analysis of access path costing may need to be performed. the optimizer will usually come up with a „good‟ plan.  If predicted cardinalities are accurate.  Bad cardinality estimates.A Simple „Methodology‟ For Query Tuning 5. when bad.

the optimizer will pick the best plan for the query. 43 . The best way of doing this is via one of the most useful tools in our tuning „toolbox‟. To start down this route of tuning we need to obtain the SQL text for the query in question along with any bind variable values it uses. the DBMS_XPLAN package.“Tuning By Cardinality Feedback”    Generally speaking if the predicated cardinalities and actual cardinalities for a row source are close.

centrexcc. this is a really easy method to use. This is also endorsed by the Oracle Real World Performance Group http://structureddata. I would not advocate playing around with density settings as this may fix queries suffering from the “Predicate Independence Assumption” but cause bad plans for other queries.pdf.“Tuning By Cardinality Feedback”      “Tuning by cardinality feedback” is borrowed from one of the reviewers of Jonathan Lewis‟s Cost Based Fundamental‟s book: www.com/Tuning by Cardinality Feedback. 44 . Other Oracle professionals may call this something different.org/2007/11/21/troubleshooting-bad-executionplans/ With the GATHER_PLAN_STATISTICS hint and “ALLSTATS LAST” format string in DBMS_XPLAN (Oracle 10g).

cpu costing is off (consider enabling it) 45 .Obtaining Your SQL Text + Binds  A trivial example to whet the appetite SQL> select * from table(dbms_xplan.:1 (NUMBER): 80310011 Predicate Information (identified by operation id): --------------------------------------------------2 .display_cursor('ft33c3agapy0k'. child number 0 ------------------------------------UPDATE TR_CYCLIC_WORK_PACK SET PACK_JOB_COUNTER =PACK_JOB_COUNTER +1 WHERE ID_PACK=:1 Plan hash value: 115273857 -------------------------------------------------------------------------| Id | Operation | Name | Rows | Bytes | Cost | -------------------------------------------------------------------------| 0 | UPDATE STATEMENT | | | | 2 | | 1 | UPDATE | TR_CYCLIC_WORK_PACK | | | | |* 2 | INDEX UNIQUE SCAN| TR_CYC_WORKPACK_PK | 1 | 9 | 1 | -------------------------------------------------------------------------Peeked Binds (identified by position): -------------------------------------1 .'TYPICAL +PEEKED_BINDS')) 2 / SQL_ID ft33c3agapy0k.0.access("ID_PACK"=:1) Note ----.

NULL.01 | 2 | ---------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------2 .cpu costing is off (consider enabling it) 46 .display_cursor(NULL.Checking Predicated Versus Estimated Cardinalities SQL> SQL> SQL> SQL> 2 3 SQL> var b1 number exec :b1 := 803100113 exec dbms_stats.gather_table_stats(user. UPDATE /*+ GATHER_PLAN_STATISTICS */ TR_CYCLIC_WORK_PACK SET PACK_JOB_COUNTER =PACK_JOB_COUNTER +1 WHERE ID_PACK=:b1 / select * from table(dbms_xplan.'ALLSTATS LAST')). SQL_ID cqnxyqmp08rtu.ESTIMATE_PERCENT=>NULL. child number 0 ------------------------------------UPDATE /*+ GATHER_PLAN_STATISTICS */ TR_CYCLIC_WORK_PACK =PACK_JOB_COUNTER +1 WHERE ID_PACK=:b1 Plan hash value: 115273857 SET PACK_JOB_COUNTER ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | ---------------------------------------------------------------------------------------------------| 1 | UPDATE | TR_CYCLIC_WORK_PACK | 1 | | 1 |00:00:00.'TR_CYCLIC_WORK_PACK'.access("ID_PACK"=:B1) Note ----.CASCADE=>TRUE).01 | 4 | |* 2 | INDEX UNIQUE SCAN| TR_CYC_WORKPACK_PK | 1 | 1 | 1 |00:00:00.

If nested loop joins are used:starts x E-Rows = A-Rows otherwise E-Rows = A-Rows   Usually for a bad plan the estimated cardinalities will differ from the predicated cardinalities by orders of magnitude.Checking Predicated Versus Estimated Cardinalities     The example on the previous slide was the simplest example that can be provided. this inaccuracy in will „ripple‟ throughout the rest of the plan and lead to poor performance. On the next slide a more complex example will be provided. 47 . Note that for line 2 the values in the E-Rows and A-Rows columns matched. Line 2 in the execution plan is what is known as a row source.

In 10g dynamic sampling or hints can help here. Not possible until 11g to gather statistics on correlated columns. more on this later. this is due to what is sometimes called “data correlation” or the “predicate independence assumption”. Not possible until 11g to gather statistics on functions unless function based indexes are used. 48 .Reasons For Estimated Cardinalities Being Inaccurate  Statistics gathering limitations:    Columns with skewed containing more than 254 distinct values => can be rectified with the DBMS_STATS API. DBMS_STATS in 10g can produce statistically incorrect estimates of distinct values when auto sampling / sampling is used.

Lack of histograms or histograms with too few buckets on columns with skewed data.Reasons For Estimated Cardinalities Being Inaccurate  None representative statistics:    Statistics are missing or stale. a particular problem with „scratch‟ data. 49 . dynamic sampling can help here. Sampled statistics taken with too small a sample size. in 11g auto sampling gives 100% statistically accurate statistics. During the course of processing data changes such that it no longer reflects the statistics in the data dictionary.

50 .Reasons For Estimated Cardinalities Being Inaccurate  Optimizer unfriendly SQL  Using values in the WHERE clauses with data types which are different to those used by the columns comparisons are being made against.

51 .Reasons For Estimated Cardinalities Being Inaccurate  Other factors: Optimizer bugs.

52 . however. the plan you have forced via hints may no longer be the „best‟ plan. #1 In the first place your stats should be correct. #2 write SQL that gives the optimizer a fighting chance of gets cardinality estimates correct. as soon as the data changes. I. #3 If you run into the “Predicate independence assumption” issue this is a tough nut to crack.e.Fixing Bad Cardinality Estimates     Hints may work fine for the data in your database at the time of testing.org/2008/03/05/there-is-no-time-like-now-to-use-dynamic-sampling/  #4 Use hints and profiles when all else fails. look at using dynamic_sampling. watch out for the use of functions and expressions. read this article first:http://structureddata.

!!! A histogram with only two end points stores no information on data distribution !!!. Only create histograms where they are required. they can have the side affect of increased hard parsing through bind peeking. 53 .Fixing Bad Cardinality Estimates     Histograms should be present for columns with skewed data with the correct number of buckets. Unless a hint is for DYNAMIC_SAMPLING only use access path hints as a last resort.

Section 3 Worked Examples Of Tuning BY “Cardinality Feedback” 01/05/2012 54 .

out_post_code || ' ' || g. f. g.address_line1. c. g.sd_customer_type.id_portfolio_address AND a.address_line11 AS best_address_line11. b.out_post_code.A „Simple‟ Example Of Bad Cardinality Estimate Trouble Shooting SELECT * FROM (SELECT a.forename. ROWNUM rnum FROM (SELECT DISTINCT a. g.address_line3 AS best_address_line3. g. b.id_servicepoint. 10011) AND sd_job_type IN (10002.ID AS customerid.address_line9 AS best_address_line9. a.address_line10 AS best_address_line10. 10003)) AND UPPER (b.id_personal_details = d.in_post_code AS best_in_post_code. a.trading_name.address_line8 AS best_address_line8.address_line5 AS best_address_line5.incomp_postcode. c. b.incomp_postcode IN ('n'. ad_personal_details d. g. sd_customer_type f. b.address_line12 AS best_address_line12.in_post_code ELSE g. g. g.in_post_code END AS out_in_postcode FROM ad_service_point a. g.address_line8.in_post_code. d. CASE WHEN b. 'N') AND a.target_read_date. b. f.id_portfolio_address = b. ad_customer c.ID AND b. c. b.better_address_flag = 'N' THEN b.address_line3.address_line9.address_line11.address_line4 AS best_address_line4. g.surname. 'WALKROUTEREFID' AS walkrouterefid.id_best_address = g. g. g. b.id_personal_details AND c.id_best_address(+) AND c.address_line5. ad_best_address g WHERE a.address_line1 AS best_address_line1. b. a. b.id_personal_details. b.out_post_code) = 'FY1' ORDER BY out_in_postcode) a WHERE ROWNUM <= 26) WHERE rnum > 0 55 .address_line6. g.address_line2.servicepoint_refid.sd_customer_type = f.id_customer.address_line2 AS best_address_line2. b. b. c.address_line6 AS best_address_line6. a.address_line4.id_customer = c. b. b.id_servicepoint NOT IN ( SELECT id_servicepoint FROM tr_job WHERE sd_job_status IN (10003.address_line7. d.*. g. ad_portfolio_address b. b.id_customer AND a. b. d. g.business_name.do_not_visit_flag. c.address_line12.out_post_code AS best_out_post_code.description.customer_ref.address_line7 AS best_address_line7.middlename.better_address_flag.out_post_code || ' ' || b.address_line10.

09 |* 20 | INDEX UNIQUE SCAN | AD_PERDET_ID_PK | 43983 | 1 | 43983 |00:00:00.access("A"."OUT_POST_CODE")='FY1' AND INTERNAL_FUNCTION("B".60 | 10 | NESTED LOOPS OUTER | | 1 | 26626 | 43983 |00:00:23.filter(ROWNUM<=26) 9 .90 | |* 4 | SORT ORDER BY STOPKEY | | 1 | 1 | 26 |00:00:38.39 ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------1 .filter("RNUM">0) 2 .11 | 18 | TABLE ACCESS BY INDEX ROWID | SD_CUSTOMER_TYPE | 43983 | 1 | 43983 |00:00:00.filter(ROWNUM<=26) 4 .A „Simple‟ Example Of Bad Cardinality Estimate Trouble Shooting ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time ---------------------------------------------------------------------------------------------------|* 1 | VIEW | | 1 | 1 | 26 |00:00:38.access("C"."ID_CUSTOMER"="C".filter((UPPER("B"."SD_CUSTOMER_TYPE"="F".22 |* 19 | INDEX UNIQUE SCAN | SD_CUSTTYPE_ID_PK | 43983 | 1 | 43983 |00:00:00.40 |* 12 | INDEX FAST FULL SCAN | AD_PORTADD_PK | 1 | 26626 | 43983 |00:00:13."ID_BEST_ADDRESS") 16 ."ID_BEST_ADDRESS"="G"."INCOMP_POSTCODE"))) 15 .90 | | 3 | VIEW | | 1 | 1 | 26 |00:00:38.36 | 7 | NESTED LOOPS | | 1 | 1 | 43983 |00:00:37.30 | 14 | TABLE ACCESS BY INDEX ROWID| AD_BEST_ADDRESS | 43983 | 1 | 825 |00:00:00."ID") | Buffers | Reads | | | | | | | | | | | | | | | | | | | | 385K| 385K| 385K| 385K| 385K| 385K| 297K| 253K| 165K| 148K| 147K| 564 90832 1 826 | 16759 | 87968 | 2 | | Writ 637 | 637 | 637 | 637 637 0 0 0 | | | | | 139K| 139K| 139K| 139K| 139K| 138K| 138K| 138K| 124K| 107K| 107K| 12M 0 | 0 | 0 | 9005K 6444K 1 16 56 .04 | 13 | TABLE ACCESS FULL | AD_SERVICE_POINT | 1 | 6762K| 6762K|00:00:20."ID_PORTFOLIO_ADDRESS") 12 .92 | 8 | NESTED LOOPS | | 1 | 1 | 43983 |00:00:37.66 | |* 9 | HASH JOIN ANTI | | 1 | 1 | 43983 |00:00:36.access("A".filter((INTERNAL_FUNCTION("SD_JOB_STATUS") AND INTERNAL_FUNCTION("SD_JOB_TYPE"))) 17 .90 | |* 2 | COUNT STOPKEY | | 1 | | 26 |00:00:38.90 | 5 | HASH UNIQUE | | 1 | 1 | 43983 |00:00:38.43 |* 17 | INDEX UNIQUE SCAN | AD_CUST_ID_PK | 43983 | 1 | 43983 |00:00:01."ID_CUSTOMER") 19 ."ID_PORTFOLIO_ADDRESS"="B".access("A".access("A".06 |* 16 | INDEX FAST FULL SCAN | TR_JOB_IX13 | 1 | 266K| 3 |00:00:04."ID_SERVICEPOINT"="ID_SERVICEPOINT") PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------11 .58 |* 11 | HASH JOIN | | 1 | 26626 | 43983 |00:00:23.75 | 6 | NESTED LOOPS | | 1 | 1 | 43983 |00:00:38.13 |* 15 | INDEX UNIQUE SCAN | AD_BESTADD_PK | 43983 | 1 | 825 |00:00:00.

A „Simple‟ Example Of Bad Cardinality Estimate Trouble Shooting   Lets create a histogram on sd_job_type Lets look at the distinct values in the sd_job_type and sd_job_status columns:- SQL> select count(distinct sd_job_type).---------------------------3 5  1 Now lets look at the histograms on these columns:- select column_name. 'SD_JOB_STATUS‘) 2* group by column_name SQL> / COLUMN_NAME COUNT(*) ---------------. count(distinct sd_job_status) from tr_job.-------SD_JOB_STATUS 5 SD_JOB_TYPE 2 57 . COUNT(DISTINCTSD_JOB_TYPE) COUNT(DISTINCTSD_JOB_STATUS) -------------------------. count(*) from user_tab_histograms where column_name in ('SD_JOB_TYPE'.

58 SQL> exec dbms_stats. but this illustrates the general principle of tuning by this approach. We now need to look at line 12 on the plan.method_opt=>'FOR COLUMNS SD_JOB_TYPE SIZE 3'. re run the query and get the estimated and actual cardinalities from the shared pool with DBMS_STATS.estimate_percent=>NULL).    .'TR_JOB'. We‟ll flush the shared pool.gather_table_stats(user. You will see that on line 13 of the plan that the estimated and actual cardinalities are closer.A „Simple‟ Example Of Bad Cardinality Estimate Trouble Shooting  The estimated cardinality is going badly wrong on line 16 of the plan.

access("A".11 | 11 | INLIST ITERATOR | | 1 | | 3 |00:00:00.06 |* 13 | INDEX RANGE SCAN | TR_JOB_IX3 | 2 | 45251 | 36201 |00:00:00.66 | 7 | NESTED LOOPS | | 1 | 1 | 43983 |00:02:36.18 |* 17 | INDEX UNIQUE SCAN | AD_PORTADD_PK | 6762K| 1 | 43983 |00:02:32.access("A"."ID_PERSONAL_DETAILS"="D".06 |* 12 | TABLE ACCESS BY INDEX ROWID| TR_JOB | 2 | 8237 | 3 |00:00:00."ID_CUSTOMER"="C"."ID_CUSTOMER") 20 ."ID") 21 .35 | 15 | TABLE ACCESS FULL | AD_BEST_ADDRESS | 1 | 102K| 102K|00:00:00.filter("RNUM">0) 2 ."ID_BEST_ADDRESS") 17 ."ID_PORTFOLIO_ADDRESS") filter((UPPER("B".00 | |* 2 | COUNT STOPKEY | | 1 | | 26 |00:03:24.00 | 5 | HASH UNIQUE | | 1 | 1 | 43983 |00:03:23.filter(("SD_JOB_STATUS"=10003 OR "SD_JOB_STATUS"=10011)) 13 ."ID_SERVICEPOINT"="ID_SERVICEPOINT") 12 .78 | 9 | NESTED LOOPS | | 1 | 1 | 43983 |00:02:29.access("C".11 |* 21 | INDEX UNIQUE SCAN | AD_PERDET_ID_PK | 43983 | 1 | 43983 |00:00:00.00 | |* 4 | SORT ORDER BY STOPKEY | | 1 | 1 | 26 |00:03:24.01 | 16 | TABLE ACCESS FULL | AD_SERVICE_POINT | 1 | 6762K| 6762K|00:00:14.access("C"."ID_PORTFOLIO_ADDRESS"="B".50 | 19 | TABLE ACCESS BY INDEX ROWID | SD_CUSTOMER_TYPE | 43983 | 1 | 43983 |00:00:00.30 |* 20 | INDEX UNIQUE SCAN | SD_CUSTTYPE_ID_PK | 43983 | 1 | 43983 |00:00:00.A „Simple‟ Example Of Bad Cardinality Estimate Trouble Shooting ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time ---------------------------------------------------------------------------------------------------|* 1 | VIEW | | 1 | 1 | 26 |00:03:24."SD_CUSTOMER_TYPE"="F".access("A".14 | 8 | NESTED LOOPS | | 1 | 1 | 43983 |00:02:35.access("A".86 | 6 | NESTED LOOPS | | 1 | 1 | 43983 |00:02:36.50 ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------1 .01 |* 14 | HASH JOIN RIGHT OUTER | | 1 | 6762K| 6762K|00:00:28."ID_PERSONAL_DETAILS") | Buffers | Reads | | | | | | | | | | | | | | | | | | | | | 13M| 13M| 13M| 13M| 13M| 13M| 13M| 13M| 13M| 91941 | 409 | 409 | 34 | 91532 | 700 | 9083 13M| 87968 2 | | W 637 | 637 | 637 | 148K| 148K| 148K| 148K| 148K| 147K| 147K| 147K| 131K| 91844 | 376 | 37 637 0 0 0 0 | | | | 0 | 91468 | 59 .76 |* 10 | HASH JOIN RIGHT ANTI | | 1 | 1 | 6762K|00:00:35."OUT_POST_CODE")='FY1' AND INTERNAL_FUNCTION("B".filter(ROWNUM<=26) 4 .access(("SD_JOB_TYPE"=10002 OR "SD_JOB_TYPE"=10003)) 14 ."INCOMP_POSTCODE"))) 18 .00 | 3 | VIEW | | 1 | 1 | 26 |00:03:24.filter(ROWNUM<=26) PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------10 .81 |* 18 | INDEX UNIQUE SCAN | AD_CUST_ID_PK | 43983 | 1 | 43983 |00:00:05."ID_BEST_ADDRESS"="G".

60 . We will use a divide and conquer strategy to work out where the predicted cardinality is going wrong.A More Complex Example    Lets turn the complexity setting up a notch. In cases where queries may take hours to run. running statements with the GATHER_PLAN_STATISTICS hint may not be practical without taking the query apart.

63 |* 21 | COUNT STOPKEY | | 1 | | 26 |00:01:31. .04 |* 15 | INDEX UNIQUE SCAN | AD_WALKROUTE_PK | 107 | 1 | 107 |00:00:00.93 | 9 | NESTED LOOPS | | 1 | 1 | 107 |00:00:02. 61 . |* 13 | INDEX FULL SCAN | AD_SERVICE_POINT_IX6 | 1 | 1 | 1109 |00:00:00.85 |* 29 | TABLE ACCESS FULL | AD_SERVICE_POINT | 1 | 5945 | 2043K|00:00:16.11 | 6 | VIEW | | 1 | 1 | 26 |00:00:02.74 | 3 | UNION-ALL | | 1 | | 52 |00:00:02.11 |* 7 | SORT UNIQUE STOPKEY | | 1 | 1 | 26 |00:00:02.63 | 24 | NESTED LOOPS | | 1 | 5945 | 1799K|00:01:30.99 |* 28 | HASH JOIN | | 1 | 5945 | 1800K|00:00:34. :---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time ---------------------------------------------------------------------------------------------------| 1 | VIEW | | 1 | 27 | 52 |00:01:33.0 |* 16 | INDEX UNIQUE SCAN | SD_FIELDREG_ID_PK | 107 | 1 | 107 |00:00:00.55 | 11 | NESTED LOOPS | | 1 | 1 | 107 |00:00:01.63 | 22 | VIEW | | 1 | 5945 | 26 |00:01:31.01 |* 17 | INDEX UNIQUE SCAN | AD_CUST_ID_PK | 107 | 1 | 107 |00:00:00. .02 |* 25 | HASH JOIN | | 1 | 5945 | 1799K|00:01:24. .6 |* 32 | INDEX UNIQUE SCAN | SD_FIELDREG_ID_PK | 1799K| 1 | 1799K|00:00:03.37 | 30 | INDEX FAST FULL SCAN | AD_CUSTOMER_IX1 | 1 | 6296K| 6296K|00:00:00.14 | 10 | NESTED LOOPS | | 1 | 1 | 107 |00:00:01.01 |* 27 | HASH JOIN | | 1 | 5945 | 1799K|00:01:20.62 | 26 | TABLE ACCESS FULL | AD_WALKROUTE | 1 | 10061 | 10061 |00:00:00. .28 | 14 | TABLE ACCESS BY INDEX ROWID| AD_WALKROUTE | 107 | 1 | 107 |00:00:00.54 |* 12 | TABLE ACCESS BY INDEX ROWID| AD_SERVICE_POINT | 1 | 1 | 107 |00:00:00.20 ---------------------------------------------------------------------------------------------------| Buffers | Reads | | | | | | | | | | | 199K| 199K| 199K| 4301 | 4301 | 4301 | 4301 4290 | 3967 | 3751 | 3749 | 304K| 304K| 304K| 3703 | 3703 | 3703 | 3692 | 3674 | 3547 3546 62937 62937 0 | 0 0 | | | | | | | | | | | 195K| 195K| 195K| 195 159K| 159K| 159K| 102K| 300K| 300K| 300K| 300K| 300K| 300K| 137K| 629 6 6 6 | continued .52 |* 18 | TABLE ACCESS BY INDEX ROWID | AD_BEST_ADDRESS | 107 | 1 | 107 |00:00:00. |* 19 | INDEX UNIQUE SCAN | AD_BESTADD_PK | 107 | 1 | 107 |00:00:00.11 | 8 | NESTED LOOPS | | 1 | 1 | 107 |00:00:02.01 |* 20 | VIEW | | 1 | 26 | 26 |00:01:31. |* 31 | INDEX FAST FULL SCAN | AD_PORTADD_PK | 1 | 67137 | 6668K|00:00:06.A „Complex‟ Example Of Bad Cardinality Estimate Trouble Shooting I said I would provide a more involved example of plan for working through bad cardinality estimate trouble shooting .74 | 2 | SORT UNIQUE | | 1 | 27 | 52 |00:01:33.63 |* 23 | SORT UNIQUE STOPKEY | | 1 | 5945 | 26 |00:01:31.11 |* 4 | VIEW | | 1 | 1 | 26 |00:00:02.11 |* 5 | COUNT STOPKEY | | 1 | | 26 |00:00:02.

access("A"."SD_SERVICEPOINT_STATUS"=10001 AND "A"."ID_PORTFOLIO_ADDRESS") 28 ."ID_WALKROUTE") 16 ."ID_WALKROUTE"="C".access("C".access("A".access("A".access("A"."ID_WALKROUTE"="C"."SD_FIELD_REGION"="ID") 17 ."ID_BEST_ADDRESS") 20 ."ID_CUSTOMER") 18 ."ID_WALKROUTE" IS NOT NULL AND UPPER("A".filter("A"."WALKROUTE_REVIEWED_FLAG")='Y' AND (("A" PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------("A".filter(ROWNUM<=26) 23 .A „Complex‟ Example Of Bad Cardinality Estimate Trouble Shooting Predicate Information (identified by operation id): --------------------------------------------------4 5 7 12 filter("RNUM">0) filter(ROWNUM<=26) filter(ROWNUM<=26) filter(("A"."SD_FIELD_REGION"="ID") Note ----."SD_SERVICE_TY 13 ."ID_PORTFOLIO_ADDRESS"="B".access("A".cpu costing is off (consider enabling it) PLAN_TABLE_OUTPUT 62 .filter(ROWNUM<=26) 25 ."SD_OCCUPIED_STATUS"=10001 AND "A".access("A".filter("RNUM">0) 21 .filter(UPPER("B"."ID_CUSTOMER"="D"."SD_SERVICEPOINT_STATUS")))) 31 ."ID_WALKROUTE") PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------27 ."ID_WALKROUTE" IS NOT NULL AND "A"."WALKR AND "A"."SD_SERVICEPOINT_STATUS"=10001 AND "A"."ID_CUSTOMER") 29 ."INCOMP_POSTCODE")='N') 32 ."ID_BEST_ADDRESS"="B"."SD_SERVICE_TYPE"=10000) OR ("A".filter(UPPER("B"."INCOMP_POSTCODE")='N') 19 ."ID_BEST_ADDRESS" IS NOT NULL) 15 .filter(("A"."ID_CUSTOMER"="D"."ID_BEST_ADDRESS" IS NULL AND UPPER("A".access("C"."SD_OCCUPIED_ INTERNAL_FUNCTION("A".

"SD_SERVICEPOINT_STATUS"=10001 AND "A".800."ID_BEST_ADDRESS" IS NULL AND UPPER("A"."SD_OCCUPIED_ INTERNAL_FUNCTION("A"."WALKR AND "A".A „Complex‟ Example Of Bad Cardinality Estimate Trouble Shooting    Start by looking for the row source which is the furthest into the execution plan for where the cardinality is out by an order of magnitude. this means that there are predicates associated with this table.filter(("A". Oracle thinks that the predicates associated with line 28 are for line 29."SD_SERVICE_TYPE"=10000) OR ("A"."SD_SERVICEPOINT_STATUS"))))   Due to a possible bug. 63 . This is line 28 in our plan:HASH JOIN | | 1 | 5945 | 1800K|00:00:34.000. lets look at these 29 . Notice the „*‟ at the beginning of the line."ID_WALKROUTE" IS NOT NULL AND "A". Now lets use a “Divide and conquer” strategy for working out where exactly in this list of predicates that are OR-ed and AND-ed together the estimate is going wrong.85 | 102K| 137K| |* 28 |  We have an estimated cardinality of 5945 versus an actual cardinality of 1.

sd_occupied_status = '10001' AND a.77 | 92450 | 92426 | |* 2 | TABLE ACCESS FULL| AD_SERVICE_POINT | 1 | 19524 | 2043K|00:00:18.id_best_address IS NULL AND UPPER (a. .sd_servicepoint_status = '10003' ) 2 3 FROM 4 WHERE 5 AND 6 AND 7 8 9 10 11 12 13* 14 / COUNT(*) ---------2043175 SQL> select * from table(dbms_xplan.walkroute_reviewed_flag) = 'Y' ( ( a.sd_servicepoint_status = '10000' OR a.42 | 92450 | 9 ---------------------------------------------------------------------------------------------------.Trouble Shooting – 1st Pass 1 SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a.sd_service_type = '10000' AND a.sd_service_type = '10001' AND a. .sd_servicepoint_status = '10001' ) OR ( a.NULL.id_best_address IS NULL UPPER (a.'ALLSTATS LAST')).sd_servicepoint_status = '10001' ) OR a.sd_service_type = '10001' AND a. PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 9qvfwxgnv1scj.walkroute_reviewed_flag) = 'Y' AND ( ( a.sd_servicepoint_status = '10000' OR a.display_cursor(NULL.sd_servicepoint_status = '10003' ) Plan hash value: 185642247 PLAN_TABLE_OUTPUT ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:17.sd_occupied_status = '10001' AND a. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE a.sd_servicepoint_status = '10001' ) OR a.sd_servicepoint_status = '10001' ) OR ( a. 64 . .sd_service_type = '10000' AND a.

sd_servicepoint_status = '10001' ) OR a.sd_servicepoint_status = '10001' ) OR ( a.walkroute_reviewed_flag) = „Y‟ and a load of stuff relating to standing data columns within brackets.id_best_address IS NULL UPPER (a.sd_servicepoint_status = '10000' OR a.walkroute_reviewed_flag) = 'Y' ( ( a.sd_service_type = '10000' AND a. 1 SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a.Analysis Of 1st Pass  We have two sets of predicates the id_best_address IS NULL and UPPER(a.sd_service_type = '10001' AND a.sd_servicepoint_status = '10003' ) 2 3 FROM 4 WHERE 5 AND 6 AND 7 8 9 10 11 12 13* 14 /  For our second pass lets work which one of these sections is causing the cardinality estimate to go wrong.sd_occupied_status = '10001' AND a. 65 .

sd_occupied_status = '10001' AND a. 66 .sd_service_type = '10001' AND a.sd_servicepoint_status = '10000' OR a.sd_service_type = '10000' AND a.id_best_address IS NULL UPPER (a. 1 SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a.sd_servicepoint_status = '10001' ) OR ( a.walkroute_reviewed_flag) = „Y‟ and a load of stuff relating to standing data columns within brackets.sd_servicepoint_status = '10001' ) OR a.Analysis Of 1st Pass  We have two sets of predicates the id_best_address IS NULL and UPPER(a.walkroute_reviewed_flag) = 'Y' ( ( a.sd_servicepoint_status = '10003' ) 2 3 FROM 4 WHERE 5 AND 6 AND 7 8 9 10 11 12 13* 14 /  For our second pass lets work which one of these sections is causing the cardinality estimate to go wrong.

id_best_address IS NULL SQL> SELECT 2 3 FROM 4 WHERE 5 / COUNT(*) ---------6712561 SQL> select * from table(dbms_xplan.filter("A".Trouble Shooting .37 | 92450 | 92426 | PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------|* 2 | TABLE ACCESS FULL| AD_SERVICE_POINT | 1 | 6712K| 6712K|00:00:13.46 | 92450 | 9 ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------2 . PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 2545fjyq0m5nm.cpu costing is off (consider enabling it) 67 . 1* select * from table(dbms_xplan.display_cursor(NULL.id_best_address IS NULL Plan hash value: 185642247 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:16.'ALLSTATS LAST')) /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a."ID_BEST_ADDRESS" IS NULL) Note ----. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE a.display_cursor(NULL.NULL.'ALLSTATS LAST')).2nd Pass  It appears that the cardinality estimate is going wrong in the section in brackets.NULL.

sd_servicepoint_status = '10000' a.access(("A".sd_servicepoint_status = '10003' Plan hash value: 2502039880 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:00. PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 4rjg28wg48zv7.03 | 6 | 5 | PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------| 2 | INLIST ITERATOR | | 1 | | 282 |00:00:00.Trouble Shooting – 3rd Pass  1 Lets break down the section in brackets.sd_servicepoint_status = '10000' OR a.sd_servicepoint_status = '10003' 2 3 FROM 4 WHERE 5* OR SQL> / COUNT(*) ---------282 SQL> select * from table(dbms_xplan.03 | 6 | ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------3 .NULL.02 | 6 | |* 3 | INDEX RANGE SCAN| AD_SERVICE_POINT_IX5 | 2 | 282 | 282 |00:00:00. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE a. SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a."SD_SERVICEPOINT_STATUS"=10003)) 5 | 5 | Note ----- 68 .display_cursor(NULL.'ALLSTATS LAST'))."SD_SERVICEPOINT_STATUS"=10000 OR "A".

access(("A".sd_servicepoint_status = '10003' 2 3 FROM 4 WHERE 5* OR SQL> / COUNT(*) ---------282 SQL> select * from table(dbms_xplan.Trouble Shooting – 4th Pass  1 Lets break down the section in brackets. SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a a.'ALLSTATS LAST'))."SD_SERVICEPOINT_STATUS"=10003)) 5 | 5 | Note ----- 69 .03 | 6 | ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------3 .03 | 6 | 5 | PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------| 2 | INLIST ITERATOR | | 1 | | 282 |00:00:00.sd_servicepoint_status = '10003' Plan hash value: 2502039880 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:00. PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 4rjg28wg48zv7.sd_servicepoint_status = '10000' a. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE a.sd_servicepoint_status = '10000' OR a.display_cursor(NULL.NULL."SD_SERVICEPOINT_STATUS"=10000 OR "A".02 | 6 | |* 3 | INDEX RANGE SCAN| AD_SERVICE_POINT_IX5 | 2 | 282 | 282 |00:00:00.

sd_occupied_status AND a.sd_service_type AND a. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE ( ( a."SD_SERVICEPOINT_STATUS"=10001 AND ("A".'ALLSTATS LAST')).sd_service_type = '10001' AND a.Trouble Shooting – 5th Pass 1 SELECT 2 3 FROM 4 WHERE 5 6 7 8* SQL> / COUNT(*) ---------2070646 SQL> select * from table(dbms_xplan.sd_occupied_status = '10001' AND a.display_cursor(NULL.sd_servicepoint_status = '10001' ) ) Plan hash value: 185642247 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a ( ( a.sd_servicepoint_status = '10001' ) OR ( a.sd_service_type AND a."SD_OCCUPIED_STATUS"=10001 AND "A".67 | 92450 | 9 ---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id): --------------------------------------------------2 ."SD_SERVICE_TYPE"=10000 OR ("A".filter(("A". PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 17vy5ysvdqgxd.NULL.sd_servicepoint_status OR ( a.90 | 92450 | 92426 | |* 2 | TABLE ACCESS FULL| AD_SERVICE_POINT | 1 | 1952K| 2070K|00:00:18."SD_SERVICE_TYPE"=10001)))) 70 .sd_service_type = '10000' AND a.sd_servicepoint_status = = = = = '10000' '10001' ) '10001' '10001' '10001' ) ) | PLAN_TABLE_OUTPUT ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:17.

71 | 18 ---------------------------------------------------------------------------------------------------- 71 .display_cursor(NULL. PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 2qd1zzbj3qgf7.Trouble Shooting – 6th Pass 1 SELECT 2 3 FROM 4 WHERE 5* SQL> / /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a ( ( a.NULL.'ALLSTATS LAST')).sd_servicepoint_status = '10001' ) ) Plan hash value: 1416185116 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:04. child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE ( ( a.sd_service_type = '10000' AND a.sd_servicepoint_status = '10001' ) ) COUNT(*) ---------1679148 SQL> select * from table(dbms_xplan.74 | 18921 | 18859 | |* 2 | INDEX FAST FULL SCAN| AD_SERVICE_POINT_IX11 | 1 | 1679K| 1679K|00:00:01.sd_service_type = '10000' AND a.

child number 0 ------------------------------------SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE ( ( a.sd_occupied_status = '10001' 6* AND a.sd_service_type = '10001' AND a. PLAN_TABLE_OUTPUT ---------------------------------------------------------------------------------------------------SQL_ID 186k84m0t9rmx.NULL.sd_occupied_status = '10001' AND a.sd_service_type = '10001' 5 AND a.Trouble Shooting – 7th Pass SELECT /*+ GATHER_PLAN_STATISTICS */ 2 COUNT(*) 3 FROM ad_service_point a 4 WHERE ( ( a.sd_servicepoint_status = '10001' ) ) Plan hash value: 185642247 ---------------------------------------------------------------------------------------------------| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers | Reads | ---------------------------------------------------------------------------------------------------| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:16.display_cursor(NULL.47 | 92450 | 92426 | |* 2 | TABLE ACCESS FULL| AD_SERVICE_POINT | 1 | 364K| 391K|00:00:18.'ALLSTATS LAST')).sd_servicepoint_status = '10001' ) ) SQL> / COUNT(*) ---------391498 SQL> select * from table(dbms_xplan.17 | 92450 | 9 ---------------------------------------------------------------------------------------------------1 72 .

sd_servicepoint_status = '10001' ) ) SELECT /*+ GATHER_PLAN_STATISTICS */ COUNT(*) FROM ad_service_point a WHERE ( ( a.Analysis Of 7th Pass It appears that the cardinality estimate goes awry when Oracle is combining the cardinalities of:1 2 3 4 5 6* and:1 SELECT 2 3 FROM 4 WHERE 5* SQL> / /*+ GATHER_PLAN_STATISTICS */ COUNT(*) ad_service_point a ( ( a.sd_occupied_status = '10001' AND a.sd_service_type = '10000' AND a.sd_servicepoint_status = '10001' ) ) 73 .sd_service_type = '10001' AND a.

which it should be by default you flush the monitoring information:- SQL> execute DBMS_STATS. this may ok for a test / development environment but it may not be practical to do this on a production environment on ad-hoc basis. last_analyzed.Analysis Of 7th Pass  To rule out statistics:  We could re-compute statistics on the AD_SERVICE_POINT table. user_tab_modifications m WHERE t. deletes FROM user_tables t. If table monitoring is enabled.table_name = m.FLUSH_DATABASE_MONITORING_INFO  Then run:- SELECT num_rows. updates.table_name AND table_name = < your table name > 74 . inserts.

last_analyzed / You can see from this that it wouldn‟t be that difficult to produce something similar that works for indexes. If table monitoring is enabled. which it should be by default you flush the monitoring information:- SQL> execute DBMS_STATS. SUM(updates).table_name. last_analyzed.Analysis Of 7th Pass  To rule out statistics:  We could re-compute statistics on the AD_SERVICE_POINT table. SUM(deletes) FROM user_tables t.FLUSH_DATABASE_MONITORING_INFO  Then run:- SELECT t. this may ok for a test / development environment but it may not be practical to do this on a production environment on ad-hoc basis. user_tab_modifications m WHERE t.table_name AND timestamp > last_analyzed AND t.table_name = m. 75 .table_name.table_name = <your table name> GROUP BY t. SUM(inserts).

Analysis Of 7th Pass  This particular issue is down to what:  people from Oracle call “data correlation” and what some people from outside of Oracle coin “the predicate independence assumption”.  The exact nature of the problem is that if there are predicates on the same table for which the data in the relevant columns is related. the optimizer always assumes that these are independent. This culminates in:=> incorrect selectivitys  => incorrect cardinality estimates => sub-optimal plans 76 .

cardinality is 1.Assume everyone in the audience knows which star sign they were born under .200 * 1/144 = 8.” “How many people will raise their hands if I ask for all the people born under Aries and in December to raise their hands? What about all the people born under Aries in March? What about the people born under Aries in April ? According to Oracle the answer will be the same for all three questions:   Selectivity (month AND star sign) = selectivity (month) * selectivity (star sign) = 1/12 * 1/12 = 1 /144 Cardinality = 1. there are 12 star signs.200 / 12 = 100. Jonathan Lewis describes this as:“ . . .Analysis Of 7th Pass  This issue has been covered in great depth by the Oracle community that focus on the CBO. . . and we assume even distribution of data – selectivity is 1/12. If I ask all the people born under Aries to raise their hands.” 77 . . I expect to see 100 hands. .5 (rounded to 8 or 9 depending on version of Oracle).

78 .  Extended statistics (Oracle 11g onwards) allows you to gather statistics on columns containing related data.  Hints to force the optimizer down the appropriate execution path.  Oracle 11g automatic tuning ?. the level of sampling performed depends on the sampling level. this causes Oracle to sample the data being queried.Analysis Of 7th Pass  Solutions to the data correlation / predicate independence assumption issue include: dynamic sampling (9i onwards).  SQL profiles (Oracle 10g onwards). Oracle uses what is known as offline optimization to sample data and partially execute the query in order to create a profile containing cardinality scaling information.

Section 4 Summary & “Wrap Up” 01/05/2012 79 .

To be affective at tuning. 80 .Summary    Having good technical knowledge will only get you so far. This will be covered on the remaining slides. your technical knowledge needs to be augmented with good practise and knowledge of common pit falls.

Tuning Objectives    Set clear and well defined performance goals. CPU / IO loading. 81 . concurrent users. timings. in terms of data volume. Remember that timings are ultimately what is important to the business. How do you know you have met your goals without targets ?.

. Etc . Do not use dynamic_sampling all over the place because I have mentioned this. Do not pick up text book or article from the internet describing a performance enhancing feature and apply it blindly across an application. Do not assume you can use the tuning advisor to solve all your ills because Oracle have mentioned this. !!! ONE SIZE DOES NOT ALWAYS FIT ALL !!! 82 . . when faced with a problem. be aware of likely causes. but appreciate that one size does not fit all.Tuning By Magic Bullets       Do not tune using „magic bullets‟.

A piece of SQL performing meaningless work can perform lots of logical reads and hence lead to a good BCHR.Tuning By Hit Ratios     Hit ratios can hide a multitude of sins. 83 .co.uk. Refer to custom hit ratio from www. Take the buffer cache hit ratio (bchr) for example.oracledba.

can you please demonstrate with an example ? (I will be using PL/SQL for coding) Thanks. Uday 84 . If yes. what is a "OK" number of the LIOs (some percentage of SGA or PIO etc ?) If possible. what steps/guidelines are to be followed. You Asked If LIOs are bad. inorder to reduce/avoid LIOs ? 3). . . . 1). is it possible to reduce the LIOs programatically ? 2).Hit Ratios According To Tom Kyte The following was posted on Ask Tom 4 December 2002 . while coding.

85 .000.000 LIO's -. their DBA's sit there and say "well. It is like the stupidest ratio of them all . over 95. It should be near 100 for most systems.000.then SO BE IT.soft parse ratio (the ratio of soft to hard parses).cache hit. . 96% -. my cache hit is 99% so all is well in the world". Massive nested loop joins are the main culprit of this.000 rows back -. tuning your SQL.000 row table and trying to get 2.000 row table to a 2. 2) the piece of advice I have for everyone: totally and forever forget the concept that "if my query ain't using an index. 1) by rewriting your SQL. However. All other ratios -. if that query could be executing without doing 1. Yet. 3) ratios STINK ( i have stronger words but this is a public forum after all ).it is very doubtful that an index should be used. If you have a query that you need to execute and the best we can do is 1.000.000 LIO's then we need to fix it. I have no ratios for you. it must be broken". . setting environmental things like sort_area_size or db_file_multi_block_read_count or the optimize_index_* parameters.000. They are Experiencing excessive LIO's due to massive nested loop joins.forget about them.000.are among the most poorly tuned systems. If you are joining a 1. There is one ratio I use -. There is no magic "good number". that is that.Hit Ratios According To Tom Kyte And we said . I have a theory that systems with high cache hit ratios.

The Fan Hit Ratio  This is a meaningful hit ratio that I first came across in a Jonathan Lewis presentation http://www.jlcomp.demon.co.uk/hit_ratio.pdf 100 * ( 1 .least (1. desired response time / actual response time ) ) 86 .

New optimizer features are constantly introduced. . .html.com/2008/03/oraclesupport-keeps-closing-my-tar. 11g will be no exception. When 11. .7 comes out.1. every feature has it‟s quirks and boundary cases under which things can break.blogspot.Tuning By Oracle Marketing      Oracle 11g is self tuning => tuning is dead !!! So why has Oracle produced an SQL test case packager in 11g => http://optimizermagic. 87 . digest and the list of bugs fixed in the optimizer . Every Oracle release has bugs and flaws.0.

88 .g.Tuning By The Internet   Some organisations publish on the internet are more interested in dominating search engine searches and advertising their books than disseminating advice grounded on well documented test cases and evidence. e. Tom Kyte and Jonathan Lewis. Prefer web sites and „experts‟ that provide worked examples.

For example. 89 . index change.Tuning By “Problem Scope”    The scope of a solution to a performance problem should match the scope of the problem. histogram than a system wide parameter change. an issue is caused by one particular SQL statement is more likely to be resolved by something such as a new index. Yes. there are DBAs out there who when faced with a performance issue will look at adjusting obscure parameters which most DBAs in their professional careers will never have any need to touch.

how it behaves under load. 90 . Refer to Cary Milsap‟s presentation on performance and skew from Oracle Open World 2008. find where the time is going to first and then apply tuning techniques accordingly. Do not blindly pick an Oracle feature conducive to good performance. Understand the “flight envelope” of you application.Tuning What You Know     Always tune what you know. pick the bottle necks off one by one.

in 11g release 1 alone we have:Extended statistics  Null aware joins  Adaptive cursors  Plan base lining  A new method of calculating density   Protect yourself from this by thorough testing 91 .Tuning Is A Moving Target  The RDBMS is constantly evolving.

contention etc Understand this basic equation and its context with your application. 92 .Understand Response Time      Response time = wait time + service time Service time = CPU time => joining. therefore that must be my problem”. work off the top waits events and drill down. parsing etc Wait time = waiting on an event I/O. Avoid blind scatter gun techniques such as looking at your software with a monitoring tool and saying “hey there is a lot of latching taking place.

when was it last serviced. you say there is a performance issue with the engine.  93 . The mechanic will probably ask:When did first start happening  Under what driving conditions does this happen  Does this happen consistently  Have there been any changes to your car. it is doing 3000 Rpm at 60 Mph.Tuning Via Flat Statistics   You get a mechanic to look at your car.

Tuning Via Flat Statistics

The moral of this story is that a single statistic in isolation is not that useful. In order to put statistics into context you need to also understand things such as:where is most of the time going in your application ?  What is the normal performance base line or “Flight envelope” for your application  Is you think you are seeing strange and anomalous behaviour, what conditions does it occur under ?

94

Know Your Tool Kit

   

Performance views 10g tuning infra structure, ASH, advisors, ADDM and the time model. SQL trace, SQL extended trace, tkprof, trcsess. O/s tools, iostat, sar etc . . . DBMS_XPLAN Your tuning arsenal should consist of more than explain plan and the ability to spot full table scans.
95

Understand Key Object Statistics

 


 

Histograms, frequency and height balanced, MetaLink note 72539.1. Index quality => clustering factors, many DBAs do not understand this, this gets a whole chapter in Jonathan Lewis‟s Cost Based Fundamentals book. MetaLink notes 39836.1 and 223117.1 Predicate selectivity, MetaLink note 68992.1 System statistics aux_stats$, MetaLink note 153761.1 Sample sizes Numbers of distinct values. Etc . . .
96

com/en/know-how-community/download-area.html “The Search For Intelligent Life In The Cost-Based Optimizer” by Anjo Kolk: http://www.Useful Internet Resources       The Oracle Real World performance group blog http:/www.com General Oracle papers.org Jonathan Lewis‟s blog http://jonathanlewis.structureddata.com/ The Oracle Optimizer Group Blog http://optimizermagic.evdbt.trivadis.blogspot. look out for those in particular from Chritian Antognini http://www.com/SearchIntelligenceCBO.com/ Wolfgang Breitling‟s web site (a reviewer of Jonathan Lewis‟s CBO book) www.centrexcc.wordpress.doc 97 .

3.91473090.:7.90/.70:80/ 89./5.3900892.-.7/3.250-0574.0 4909.0.42500..90/.908 %00.0 3903098/0.908/1107174290570/..7/3./0/ 3039000./0/  .0/ 130890/445438.798 #48 #48 490780  #48 #48 &8:.7/3.4:88/0.90/./9454475071472..98343..2504390570.3 7550 974:4:990708941905..8908250890.147.38.39:/0 983..3..3/ #48.908 -47/078412.4:2382.3!70/.7484:7..3/ 0.3-0574.:9435.9.8.9.:08390 #48..90/ '078:8892.2509.2470.

3.90/ ..:7..8.089.:08..907329.989.3-070.9 .90 $9..#0.908033.4770.7/3.:0803.3324709.9438 4:2389800/.3574/:.910/990$*$%%$! $*$%%$3 ..9 0892..439.:948.90841/893.989..843847892.9.253.3/893.

88:25943 2470 4398.8431:3.80/3/008.843. .943 4790 570/..9438:3088 1:3.4770.0.305070 495488-0:3994.0/ /.989..4:238 988/:094.903/0503/03.90789.907 3 /3.2.70:80/  .943-.25347398.989.9.8..9884209208.4770.2538 :80/ 495488-0:3994.8.90789.90/ .

.8  .7./.#0.8.7028834789.:7.9.989.5. /.09843 .7/3.7 574-029 8.9.305070 .843847892.418947.9.7 .943.08 89....4:2389800//.79.9.:948.90/ .:.253.28994410-:.8 $9.9.9.89..0 :7390.0883/.989.2.28478947..:7. $.908033.. /3..3088:.03994482.4:78041574.989.989089.8.089.9089..989.989..8.90 430705708039.989.253.9..9934 43077010.8390/.25080 3 .250/89.

#0.70/1107039949480:80/-90.9.389  .70-032.....:8089/.78438...:08390#./0.425.4:238 .90/ .843847892.908033.:7.90 59207:31703/$" &83.9508 .7/3.

..:7.90/ .908033..7/3.9478 59207-:8  .#0.90 9071.843847892.

9.9.9892.80.0.0/.9884:/-0. 44.9.9 790$"9..*8..3.308 905.9  0 .88443.0890459207..88:25943  88:0988.908 3982.4713014790/.79.903/0503/03.890/.9438.3/057088438 14:7:339490 !70/...253 70.990920 4190893 40.9./.908..01789 995.04:789..7..344307-090 -089 5.9:83 /3.4:914790:80 411:3.4770.041 098.3 4:.2.3.90892.4770.-.07 .7/3.3982..9.193.0147.3 39017895.34:7/./98.7/3.94:3:994.

.

9:70//.9. 47.897:.

 .

.

.

8.9070 8 34 920 0 34 94 :80 /3.253.2.

&80398.0801.8  .3/57410803.

7/3.4:2389 800//.110.89708479  .783974:-3/5003 8947.80/.9..29439403/54398894708 3431472.9398.9892.3.098 3..94343/.70706:70/  90.908 8947.398147*$!43:80 .4770././897-:943 &3088.990.93:2-0741-:..70..7/ 5.9413.8.3.2807090.0908/0.0885.2884:/-05708039147.9.908947.70..

2508 1%:33 . .943 470/.7/3.$0..900/-.

.

   .

//7088 #.//7088- /*5479144*.4/0 # .//7088*30  .709*70. /*5479144*.//7088*30 - .74:90701/ .8 .4/0  - 3*5489*.4/0 - -09907*.//7088*30 - .0*5439.:894207.//7088*30$-089*./.05439 # 97*4#8/*4-*89.//7088.//7088*30 - ./*807.89*1.250 1. /*507843..20 - 4:9*5489*.:894207*701 .20 / 2//03.//7088*30  .4/0 .8/ /*507843.20 .4/0   3*5489*./3*3.//7088*30 - .//7088 ..//7088*30  . .05439*701/ . .90 .//7088*30$-089*.90%74:-0$4493 $% #  $%.//7088*30$-089*.05439 % $%/*807. /*-089*.//7088*30  .4/0  3*5489*.//7088*30  . 8/*. .//7088*30$-089*.  %- 4:9*5489*.//7088*30  .//7088*30$-089*.//7088 /*-089*.//7088*30$-089*.:894207 1 $..4/0 $4:9*3*5489.//7088*1../*/.8 . /*807.:894207*9501 .//7088*30 - .425*5489. -:83088*3./*507843. /*507843.05439  # &%# $. # &73:2 #  $%$%%.//7088*30  .//7088*30  .//7088 .4/0$-089*3*5489*.4/0 3   .//7088*30$-089*. ## & #73:2  .//7088*30 - . 8/*. /*807.:894207.4/0$-089*4:9*5489*.9:8     8/*4-*950     &!!# - 4:9*5489*.20  / 8:73.7/3. /*.//7088*30 - .//7088*30 $-089*.:894207*9501  - 3.//7088*30$-089*.//7088*30  . 9.4/0  - 3*5489*.75943 .4/0 - 3.//7088*30  . .4/0 $ 4:9*5489*.*/09.*/09./*5479144*. /4*349*./*.//7088*30 1 /08.8/ 8/*. 97.//7088*30$-089*.//7088*30   4:9*5489*.//7088*1.//7088*30  - .*/09.//7088*30 - .20 / 14703. - .//7088*30$-089*.:894207/ $- -09907*.:894207 . /*.//7088*30 - .:894207*950 .4/0   ##4:9*3*5489. /*.//7088*30  - .*/09.9 892.425*5489./*-089*.//7088*30$-089*.. $250 ..//7088*30  .4/0  . 807.

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

4:23*3.9.20 $".990/893.9908947.9:8.98/*4-*950 .9:8 174297*4-  &% $%%$* *%!  &% $%%$* *$%%&$  409844.4:23*3.28439080.90%74:-0$4493 098.4:39 1742:807*9.4:238 $"800.70.7/3.98/*4-*89.203 $* *%!  $* *$%%&$  74:5-.4:238 800.20 . $250 ./.2438/*4-*950 09844.250 1.9.4:39 /893.90.8947..9 892.:083908/*4-*950.9.4:39 /893.3/ 8/*4-*89.-*8947.4:23*3.28070.

 &* &% $* *$%%&$ $* *%!  .

250 1.98 :807 %#*  $* *%!$ 0892.9:. $250 ..50419:33-98.039&  2094/*459  # &$ 0 1:8908.-0*89.908900307.93043905.3 -:998 :897.90/.70/5449 $*$%%$ 4:8009.9:..3/.70.3/0990 0892..7/3.573.  .90/ ../-28*89..7/3.907*9.39./743433041 905.9900892.90*507.9081742908..5574..7/3.9 892.9843-.90%74:-0$4493 %00892.7/3.4807 034300/9444..3 $"00.90/.908./.70/544 707:3906:07.3/.9433041905.98 .

7/3./...943 /03910/-4507..943.088  *&$% # *&$% #   ...088 $* *%!  #$* *%!   . $250 .90%74:-0$4493 / 507.088  *! #%  *#$$ *! #%  *#$$ 1907 &!!#   &%*! $%*    %#*&%    !*! $%   .943/   1907 #&  1907 # &  1907 # & !*%* &%!&%   ...798 #48 #48 %20:11078#0.088  $*&$% #*%!   .088  *$#'! %*$#'! %  1907 $* *$%%&$  #$* *$%%&$   ...9031472.20$9.9 892..088  *$%*#$$ *$%*#$$  ..088  *!#$ *%$ *!#$ *%$           ...250 1../8  '       &%$% !    '      $ #% ##$% !    $&"&    $% !$    $% !$    $% !$    $% !$      $ #%%     $%%#% #       %$$# %#*       #$%#* *      $ #% &%#     %$$&*$%*#$$       %$$&*$#'*! %      &"&$*! #%*!     &"&$*&$%**!      %$$# $*&$% #*%!       &"&$$*&$%%!**!     &"&$*!#%**!     !70/..

4250980993:5..349.7/3. 0:80.9020398990 %#*!*$%%$%$392.436:07897./0.9843743 3.8080706:07082./.3/.9.349-0 57.3906:07.909447 4:907090570/.94:99.4704250.5.04:78947:3  7:33389.9..90/.79  ..250 0989:7390..

90974:-084493    / 507. 4250 .90%74:-0$4493 8.798 #48 #48 %20:11078#0.4.90892./8 '     $ #%&"&     &         '           &%$% !         '          $ #%&"&$% !      $% !$        $% !$        $% !$      $% !$       %$$# *$#'*! %     &$*$#'*! %*     %$$# *# &%       &"&$*# &%*!      &"&$$*#**!       &"&$*&$%**!       %$$# *$%*#$$      &"&$*$%*!        '       &%$% !     '      $ #%&"&$% !    $% !$      $      %$$&*# &%       $       $        %$$&*$#'*! %      $%&$*&$% #*    $%&$*! #%*!      &"&$$*#**!      .20$9./ .24703.0/0.3147473974:-./4:/574..7/3.943.4393:0/    ./0.250 1./ .7/3.9892.250415.

..088  *# &% *# &% !*%* &%!&%  .250 1...90%74:-0$4493 !70/.943/      1907 #& 1907 # & 1907 # & 1907  *# &%$ %&&!!#  # &%*#'*     !*%* &%!&%  $*$#'! %*$%%&$  $* &!*$%%&$  $*$#'*%  1907  *$%*#$$$ %&  .088  $**#  490 .48938411 .088  *&$% # *&$% #  1907  *# &%$ %& *$%*#$$$&&!!#  #  $*$#'*%!  #  $*$#'! %*$%%&$  $* &!* %#*&%   $*$#'! %*$%%&$  1907 &!!#   !*! $%     .9892.088  $**#   ./ .943 /03910/-4507...088  *$%*#$$ *$%*#$$   1907 #&  1907 # &  1907 # &  .....7/3.438/0703..088  *# &% *# &%  .5:...088  *&$% # *&$% #  1907 &!!#   !*! $%     . 4250 ..9031472.088  *! #%  *#$$ *! #%  *#$$  ..-39 !*%* &%!&%  ...

.90843743  .93988941570/.941   49.8901:7908939490 00.99080  1907  *# &%$ %& *$%*#$$$&&!!#  #  $*$#'*%!  #  $*$#'! %*$%%&$  $* &!* %#*&%   $*$#'! %*$%%&$ :094.3/..7/3...9:.884./0.90/9989.7014730 4098:80..70 # 0/.:9435.3..0.90/..99070..990-0333419030 9820.984:9-.389.70 570/.7/3.7/3.7/3. 4250 .9892.30892..3  $       0.908.884. .090 .-0 09844..347/07412.436:07 897..79-443147907484:7.9089.901474734:9070 0.908..9.90/9 30.078:8.3/ 0/94090790 0892.90%74:-0$4493 $9.314707090..990570/.5488-0-:  7.39:/0 %883034:75./ .09389.0.941.250 1.

88 $%.%74:-0$4493 89 !.

%#*!*$%%$%$ .

05439*89.. 8/*807.05439*89. 8/*807.9:8     # . /*-089*...9:8   #..05439*89. 8/*807.     .:50/*89.00/*1. . 8/*4.74:90*70.  &% # .9:8     .9:8    .0*950  .9:8     #.//7088$& &!!# ../*807... 8/*807. 8/*807.0*950   .05439*89. 8/*807. #..0*5439.

13.3 /85.9 17429.8.-0 /-28*5.  &%   $"800./3:2-07 $%.*. .:7847 & & $%%$$%  !*%* &%!&% $"*6.

%#*!*$%%$%$ .

20$9. 8/*4.05439*89...0*950   .8.9:8  #. 8/*807. 8/*807.9:8   . 8/*807.9:8    !. .0*950  .# .0*5439..3.943... /*-089*.:50/*89.05439*89..05439*89.:0 !*%* &%!&% / 507..05439*89. 8/*807./*807.798 #48 #48 %20:11078#0.74:90*70./8 $ #%#%      %$$&*$#'*! %          .//7088$&&!!# . 8/*807. 8/*807.00/*1.     .9:8     #.. &% # .9:8    # ..

9..4:23893-7.88 189 !.3/ &!!# .098 $%.   . ..//7088$&../4189:1170.3.3/.00/*1.4.3/3/.094809841570/..88 0.74:90*70.90890/*-089*.9394 89.

%#*!*$%%$%$ .

8/*807.05439*89.9:8     # . 8/*807. 8/*807.//7088$& &!!# ..9:8     . 8/*807.. 8/*807. 8/*4. /*-089*.0*950   ...9:8    . .9:8     #.0*950  ./*807.74:90*70.0*5439.. #.  &% # ..:50/*89.05439*89.00/*1.05439*89.9:8   #..     ..05439*89. 8/*807.

474:780.8809847.43/5.7/3.94388.:83 90..90944743  .90892.43041908080..

.74:90*70.3./4189:1170.. .4.3/ &!!# .098 $%.3/3/.88 189 !.9.//7088$&..88 0.4:23893-7.   ..094809841570/.3/.00/*1.9394 89.90890/*-089*.

%#*!*$%%$%$ .

8/*807....9:8    . 8/*807.9:8   #.9:8     #..     ..//7088$& &!!# .:50/*89.  &% # ...0*5439.0*950  .00/*1. .05439*89.9:8     # . /*-089*. 8/*807. 8/*807.74:90*70./*807.05439*89. 8/*807.9:8     . 8/*4.05439*89.. #. 8/*807.05439*89.0*950   .

.90892.7/3.43041908080.:83 90.43/5.94388.8809847..90944743  . 474:780.

88 9.990.:7847 & & $"$%.3 /85.7/3.%74:-0$4493 3/ !.-0 /-28*5.098  800.9084374339080.550..9 17429.90892.9433-7..*.789.

%#*!*$%%$%$ .

#.//7088$& ..0*5439. /*-089*.  &% # ./*807.

 &%  $"800.-0 /-28*5./3:2-07 $%.*.3 /85.:7847 & & $%%$$%  $%%$$% !*%* &%!&% $"*16 232 .9 17429.

%#*!*$%%$%$ .

//7088$& !.9031472.# .-39  . /*-089*. &% # .943 /03910/-4507..798 #48 #48 %20:11078#0..3./*807.48938411 .8..943/   1907  *$%*#$$$& 490 ./8 $ #%#%     !*%* &%!&%  %$$&*$#'*! %     !70/.943.438/0703.5:.20$9.:0 / 507.0*5439.

9433-7.%74:-0$4493 7/ !..88 098-70.098 $%./439080.

%#*!*$%%$%$ .

.  &% # .05439*89.9:8   $".9:8    #. #.0*5439...05439*89./*807. 8/*807. 8/*807.

:7847 & & $%%$$%  !*%* &%!&% $"*7.3 /85.*./3:2-07 $%.  &%  $"800. .9 17429.-0 /-28*5.

%#*!*$%%$%$ .

05439*89.9:8   #.8...798 #48 #48 %20:11078#0. 8/*807..943/   .088 490  $*$#'! %*$%%&$  # $*$#'! %*$%%&$   ..0*5439.9031472.943 /03910/-4507./8 $ #%#%    !*%* &%!&% $%%#% #     #$*$#'*! %*    !70/. 8/*807.05439*89. &% # ./*807..943..# ..3.20$9.9:8   !.:0   / 507.

098 $%..9433-7.%74:-0$4493 9 !.88 098-70./439080.

%#*!*$%%$%$ .

/*807.0*5439.05439*89. 8/*807.9:8    #..05439*89. #. 8/*807.9:8   $"..  &% # ..

 ./3:2-07 $%.3 /85.9 17429.*.  &%  $"800.-0 /-28*5.:7847 & & $%%$$%  !*%* &%!&% $"*7.

%#*!*$%%$%$ .

. 8/*807.8./*807.943/   .9031472./8 $ #%#%    !*%* &%!&% $%%#% #     #$*$#'*! %*    !70/.9:8   #.943.:0   / 507....9:8   !.088 490  $*$#'! %*$%%&$  # $*$#'! %*$%%&$   .# .3.. 8/*807.20$9.0*5439.05439*89.798 #48 #48 %20:11078#0..05439*89.. &% # .943 /03910/-4507.

%74:-0$4493 9 !.88 $%.

%#*!*$%%$%$ .

.05439*89...9:8  . 8/*807..  &% # .. 8/*807.9:8  # .0*950 ..05439*89. 8/*4.9:8 $". #  .0*950 . 8/*807.0*5439. 8/*807./*807.:50/*89.

 &%    $"800./6/ .8.-0 /-28*5.9 17429./3:2-07 $%.3 /85.:7847 & & $%%$$%              !*%* &%!&% $"*.*.

%#*!*$%%$%$ .

943 /03910/-4507.:50/*89.05439*89. 8/*4.0*5439.20$9.9031472. 8/*807.. 8/*807.3./*807.9:8   .:0 / 507.. &% # .0*950   . 8/*807.943/   1907  $*$#'! %*$%%&$   $*$#'*%!  #  $* &!*$%%&$  $*$#'*%!   ..943../8 !*%* &%!&% $ #%#%       %$$&*$#'*! %       !70/.#  . 8/*807.9:8    # .798 #48 #48 %20:11078#0.8....0*950  .9:8     !..05439*89.

%74:-0$4493 9 !.88 $%.

%#*!*$%%$%$ .

0*950   .. 8/*807.9:8     $". 8/*807..05439*89./*807.  &% # .0*5439.. #  .

:7847 & & $%%$$%  !*%* &%!&% $"*6/-61 ./3:2-07 $%.9 17429.3 /85.-0 /-28*5.*.  &%  $"800.

%#*!*$%%$%$ .

05439*89.0*5439.798 #48 #48 %20:11078#0...8. &% # .0*950  .:0 / 507...9:8     !. 8/*807.20$9./8 $ #%#%      $%&$*$#'*! %*      .3./*807. 8/*807.943.#  .

88 $%.%74:-0$4493 9 !.

%#*!*$%%$%$ .

0*5439.0*950   .9:8     $"..05439*89.  &% # .:50/*89. #  ..9:8    . 8/*807./*807. 8/*4... 8/*807.

*.3 /85./3:2-07 $%.:7847 & & $%%$$%  !*%* &%!&% $"*2 972 .-0 /-28*5.9 17429.  &%  $"800.

%#*!*$%%$%$ .

9:8     !./*807.8. 8/*807..:0 / 507.05439*89.3.0*5439.798 #48 #48 %20:11078#0. 8/*4.943...:50/*89..0*950   .9:8   . &% # .20$9.. 8/*807./8 $ #%#%      %$$&*$#'*! %      .#  .

3.88 19 !..7/3.88 9.42-3390.08.7/3.990.90892.550.703 7.90408.789...90841 $%.

%#*!*$%%$%$ .

9:8    .9:8     .0*5439. 8/*4. 8/*807..3/ $%.... 8/*807. #  .0*950   ./*807.  &% # .05439*89.:50/*89.

%#*!*$%%$%$ .

..0*950   .0*5439..05439*89.  &% # . 8/*807. 8/*807./*807.9:8     $". #  .

 .

3.989.9089.84390*$#'*! %9.8 0.4147.88 %47:04:989.88 19 !.4:/70 .425:9089.989.-0 982.

.984:/-0-/01.7432039-:992..-0/ .943 $"00.7432039 43.-0*3.20 9.-0*3.3.-089 :807*9.-.94303.202 9.349-057.-0*3.:94:1:8902439473 31472.94/49843.0/ 380798 :5/.9.:90$*$%%$ &$*%$* % #* %037:3 $%3:2*748 . /0.045203903./ 4.908 /00908 # :807*9.88 19.204:79.574/:.-03.89*..94382 #9 9.-02439473803.-*24/1.20  .

989.4147.9089.88 %47:04:989.8 0.989.4:/70 .84390*$#'*! %9.-0 982.3.88 19 !.425:9089.

204:79.9.-02439473803.20 # &!9 9.0/ $& 380798 $& :5/.7432039 43.574/:.20 92089.20 .89*. /0.88 19.-0*3.20 .94382 #9 9..-0*3.984:/-0-/01.045203903.-0*3.3.-089 :807*9.7432039-:992..0/ 9 9.94/49843.25.-03.:94:1:8902439473 31472.0/ .349-057.3.:90$*$%%$ &$*%$* % #* %037:3 $%9 9.-0*3.89*../ 4.94303.202 9.-0*3.-.-0/ .3.-*24/1.943 $"00.89*.908 $& /00908 # :807*9.

9 4781473/008  .084209382. 4:.79..:994574/:.994:/3 9-09.38001742989.9/11.

90892.0 .4770.943 .90/ 90459207..4770.209.99080.703/0503/039 %8.3.4:238870.903/0503/03.88 19 !.88 %85.7/3.9.:23..9.98 3.88:2089.88:25943 %00.39.984205045017424:98/041 7.919070..9083 3..3/.0.9 504501742 7.43 90570/.9:704190574-0289.79.788:08/4394.9. /...90843908.8 .9..5.4770..:.93.9800.38  .0.-0 147.390700.908 8:- 4592..70570/.90/..

07430390.:/03.8.3/0.78390070-473 :3/07   1.9..:8 4390 43.994 800 .9.0348.89. 800.9/059-90 7.8    88:200.914.88:200.7-0898.88 %888:0.8-003.9050450-473:3/07708947.308/08.80907.9.3/8 050.0.03/897-:94341/.4.88 19 !.7838 .7089.3.3/8 9070.422:399.98.070/370.

. .98  .7/3.

9050450-473 :3/077083.02-07947.8147.-4:99050450-473:3/07708357 . 42.9 89..9.783  .97006:089438 $00.9.80907.3/30..783 800.9.9 243989.3/81.3807-0908.47/394 7.090.3/8.80907.20147.9 2439  800.9050450-473:3/07 708.3504507.9.9.-4:9..7.

 .

.

9   . .7/3.

.0784341 7.0    .  74:3/0/9447/0503/343.

88 19 !.943.88 $4:94389490/.3..9.4770.

33.3/5.8 411304592.570/.439.4:238.88:25943 88:03.90//..7/8  7..043.0 43.57410.:9435.7/3.943 903/0/89.903/0503/03.9:33  .0.9.0.250/.3370..25090 /.7/8 .:/0 /3.8.943948.0948.557457.90.9.:90906:073 47/0794..439.70..989..9000.090459207/4390...98343.9.2530.98.7/8 98.:808 7.8 7.9 $"574108 7.989.79.9.25350714720//0503/84390 8.00...907 89.0:808.:942.2.331472.0 39894147.-036:070/ 900.253 43.0418.484:94. 7.843.

5&5 .943 $:22.7  7.$0.

.

   .

$:22.9..340/043094: 841.0.3.070/4390702.99:33 4:790.:20390/944/57.8 %8-0.338/08  .340/0 300/894-0.4.110..42243591.3/ 340/041..7 %4-0.7 .3.980.344/90..

9.9.04.08 $09.3/0/0130/5071472.0.8 3 9072841/.7.%:33 -0.:77039:8078 9238  !&..43.4:20 .3.

98 25479..70:92.02094:74.399490-:83088  ./3 4/44:344:.7098 #0202-079.99238.894:9 9. 4.90.

0.79..559 -3/...3. 574-02 -0.0-0..02039430/98 4349:80/3.:80 7.*8.003.9430 80/40834919..3.310..7-3..0.2.:80 .3.-:098 031.7488.07905.4.%:33.3/..253...0/9.01742903907309 /08..943 4349.9:70.3:80909:33.02039430/98 9. 43495.5570.88:204:...5071472.:808 -:9..909.:5909-4447../.55. 4:78-0.    $ $ %$%  .70410.8479484.:098 43499:30:83 2.

70.0/-.3308847.941742  47..4 :  .00..94 -./94.. 44/# #010794.:894297.2:99:/041838 %...097../8. .250 50.3/03.%:339#.948.7 1470.090-:1107.948 97.041$"5071472320.3 5071472498414.3/0.

70-.948.47/3%4%4290 %01443.02-07     4:80/ 1 8..2.  108 .090 85747./   895488-09470/:..989058.9.9#.854890/438%420.

4/3 347/0794 70/:.:/0308.0.7094-01440/ 0.

30. 15488-0 .98.250 -0:83 !.041$47!  09.909.039...4/ 8  .. 3:2-074190 8 8420507.34:50.80/0243897.

4/3 %.  .38 &/.$"147 .

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

.39#.91789.20. 43.331:97.3085708039..94 %88.%0.9.74883.949.943 995.20.

.

 .4 :.425 /0243 .

94 5/1     0.9*7.89  /0870/70854380920.

.70854380920  . .9:.

%:33 7.0/.073 995..805.088019:339:338/0.7093 7.0574/:...3$"9089./ $4...0.8 7..

.

.42.4592072. -48549 .

 .

.

8-:8.3/908941-:810/3 90459207    045920710.  .70.9:70. 938.393974/:..8 -034 0.808:3/07.483 2 9.89 86:78.3-70.3/-4:3/.0700.3/1.05943 03  .9:708.0 8:55479 0058 .07 7.80.0/ 0...7.42084:9 /089.7 92 .4389.07 10.47..

70 247039070890/3/423.3/.:20390/9089./.07983907-4489.3 /88023.3/43.2508 0  %4290.808.08.0330 80.%:33%03907309 $42047.7.9380.3/0..7.074:3/0/430 /4.9.94385:-843903907309../0470/ 0./03./.3/ 050798 9.308  .9574.0 !701070-8908.38.93.

88902 /05..3.70078 30.45041.908.9.5071472.388:08..20907.84:94394.29. 5071472.250 .7084:990704031.450 %08.3./:8934-8.209078.0574-0284:/ 2.:.90203982470094-07084.248983907574108843.7..0/9.8.  .. ..79.3300/9494:.:70 5.07.%:33 !74-02$.0/-8420938:.3.4504190574-02 470.7.30 08 9070.7$" 89.9.:80/-4305..088:044.30 8947..303/0 3/0.0.

3.0.3/90 1903.5 85708039.47/3 &3/0789.3./ #010794..0450 414:..%:33.94:34 13/0709092084394 1789.55.90-499030.3/ 801742 7.8411430-430  ..43/:.0 50347/  4349-3/5.010..89:30.943 49 -0.3/903.9:70..3 7.0 5.94:34 .559:3390.36:08.78..09444/ 5071472.943435071472...08:3/074.

3 3700.:78478 !.0.80 .3-...4389.989.0 903/0/89.%:3384.4.8033 302094/41.94:7801174298-9474:90893  .4300.3%..59.70438 /.709 %0#$8.93/0389 !7490.:.8 :..390.

9920 807.9920.&3/0789.9343.30.039.0920!&920433 5.3/#0854380%20 #0854380920. .78309.0920 $07...

&3/0789.0 907014709.3/98.9907:390.3/8. .36:08 8:.06:.4390994:7 .2439473944.8443.3/98-.4/-3/8.3//7/43  .92:89-02 574-02 4741190945.39..4941.8.55.709.943 ...9 4:78419.35.943.3 090708 .9..4390394309.0398.980.

%:33'.-..8 03//178989.7 4:8.989..3.9/7.0/  .79.8 4:09...43889039 .3.3..89.5503 40898.9 5 %020.5503.574-.43/9438/40898.89807.308944:7..3.94:7..55033 &3/07.09070-003.9$9. 90708.7 03 .088:09900330 98 /43 #52.3.20..9444.5071472.

439094:300/94 .30.98903472..943 84:934:.5071472.3.9.:7:3/07  .4:7 .708003897.989.9:801: 347/07945:989.0-.0450 1474:7.989.94383499.9..43/9438/40894.943 .9$9.84:3/0789.803047 9 03..4198894789.8394.55.8 %0247.8 0708248941909204334:7.%:33'.55..3/..342.4:8 -0..83089.3/9388:.989.3 84.

9:70 $ .3/9092024/0 $"97.08  9:33317.3.8478  ./..344:7%449 !071472.897:.8088 .0 $"0903/0/97.0 95741 97..0.

3/90.38  .43889412470 9.84:/.35.3.709.9 8.7803.-08.   $*! 4:79:33..89448 489.-99485491: 9.305.

98 09.989.5907343.33490  $890289..:*89.8 8947.33490  3/06:.33490  $.9.38/4349 :3/0789.250808 :2-07841/893.9$9.989.9478 2..    ..0/ 09.40.8-44 09.:08 9.80/:3/.9.&3/0789.334908 .3/0 -0..3 08 8489.9 09.9.2039.28 1706:03...90800.:890731.3.3/98 98098.9.8.3/09-.3/  !70/.

08 %0 7.0#0.&801:3907309#084:7.074:5-4 995..47/5071472.3.

9.9.308 8-4995. 47 43. 897:.9:70//.

.

43.9.42.308 47/57088 .

0 5920774:54995. %0 7..

.

4592072. -48549 .42..

7174279.5078 444:9147948035.42 0307. ..03970. 41.0074143.79.:.308 8 -44   .05.3 39433 995. 7..70.37093 80-890 .9.

.

./8 .42. 97.

03.

34 4 .422:39.

7./434. 92 %0$0.47390039103%0489 ./ .80/ 59207 -34 4995.70.

.

42. 0./-9 .

 .$0.0 /4.39003.7.