Wednesday, 4 March 2020

Best experience: Creation of HANA Indexes part II

Content:

1. Starting with indexes (part I)
2. Limitations (part I)
3. SAP recommendations (part I)
4. Evaluate expensive statements (part I)
5. Analyzing the system
6. Performance comparison
7. Tools

5. Analyzing the system


If you don’t have any performance issues and no complaints from business users stop optimizing a system which don’t needs it. Every index has also side effects which should be evaluated carefully.

For the start check the current primary key and existing indexes of your table.

This can be done via scripts from the SQL collection (note 1969700).

If you only have the table name use the following statement:

HANA_Tables_ColumnStore_Columns

( SELECT                                      /* Modification section */
          '%' HOST,
          '%' PORT,
          '%' SERVICE_NAME,
          '%' SCHEMA_NAME,
          '<insert_table_name>' TABLE_NAME,
          '%' DATA_TYPE,
          40  COLUMN_NAME_LENGTH_LIMIT,
          '% ATTRIBUTE_TYPE,       /* CONCAT_ATTRIBUTE, TREX_EXTERNAL_KEY, TREX_UDIV, ROWID, TEXT, UNKNOWN, % */
          '%' COMPRESSION_TYPE,
          '%' LOADED,
          '%' PAGEABLE,
          ' ' ONLY_INTERNAL_COLUMNS,
          ' ' EXCLUDE_PK_AND_UNIQUE,
          ' ' GENERATE_DROP_CONCAT_ATT_COMMAND,
          'TABLE' OBJECT_LEVEL,    /* TABLE, PARTITION */
          'NONE' AGGREGATE_BY,     /* HOST, PORT, SERVICE, SCHEMA, TABLE, COLUMN, DATA_TYPE, ATT_TYPE, PAGEABLE, LOADED, COMPRESSION or combinations, NONE for no aggregation */
          'SIZE' ORDER_BY,                          /* SIZE, TABLE */

Please check the compression type (see part I – sparse / prefixed) for the possible affected index columns.

If you already have the statement hash use (which also includes the CS columns):

HANA_SQL_StatementHash_DataCollector* (depending on your revision)

( SELECT                /* Modification section */
      '1000/10/18 07:58:00' BEGIN_TIME,                  /* YYYY/MM/DD HH24:MI:SS timestamp, C, C-S<seconds>, C-M<minutes>, C-H<hours>, C-D<days>, C-W<weeks>, E-S<seconds>, E-M<minutes>, E-H<hours>, E-D<days>, E-W<weeks>, MIN */
      '9999/10/18 08:05:00' END_TIME,                    /* YYYY/MM/DD HH24:MI:SS timestamp, C, C-S<seconds>, C-M<minutes>, C-H<hours>, C-D<days>, C-W<weeks>, B+S<seconds>, B+M<minutes>, B+H<hours>, B+D<days>, B+W<weeks>, MAX */
      '<insert_hash>' STATEMENT_HASH,
      -1 PLAN_ID,
      20 MAX_RESULT_LINES,
      86400 TRACE_HISTORY_S,
      200 LINE_LENGTH,
      ' ' SHOW_COMPLETE_BIND_VALUE_LIST,
      'H' TIME_UNIT                    /* MS, S, M, H, D */

In our example: Hash 426bdc1658f5f21d5bebf85ec4ab9c2c

******************
* THREAD SAMPLES *
******************
SAMPLES PERCENT HOST     PORT  THREAD_TYPE
======= ======= ======== ===== ==========
   5033   54.07 hostname 30040 SqlExecutor
   4275   45.92 hostname 30040 JobWorker
SAMPLES PERCENT THREAD_STATE     LOCK_NAME                                                         
======= ======= ================
   8209   88.19 Running          ?
   1093   11.74 Job Exec Waiting JobBarrier Wait for Jobs
      3    0.03 Job Exec Waiting ?
      2    0.02 Mutex Wait       unnamed Mutex
      1    0.01 Mutex Wait       ?
SAMPLES PERCENT THREAD_TYPE THREAD_METHOD                                 
======= ======= ===========
   5033   54.07 SqlExecutor ExecutePrepared
   4043   43.43 JobWorker   RleScanVecOutJob<range>
    230    2.47 JobWorker   <unknown method>
      2    0.02 JobWorker   BarrierEmptyJob
SAMPLES PERCENT THREAD_TYPE THREAD_METHOD           THREAD_DETAIL                                                               
======= ======= =========== ============
   5033   54.07 SqlExecutor ExecutePrepared         SELECT COUNT(*) FROM "VBEP" WHERE "MANDT" = ? AND "VBELN" = ? AND "EDATU" = ?
   4043   43.43 JobWorker   RleScanVecOutJob<range> ?
    230    2.47 JobWorker   <unknown method>        <unknown job>
      2    0.02 JobWorker   BarrierEmptyJob         ParallelDispatcher

ExecutePrepared stands just for the execution of an already prepared statement which is normal and no indicator for an intervention.

Completely different with RleScanVecOutJob<range>. This thread method is a definitely an indicator for a need of an index.

Another indicator can be check 890 and 1125 of the mini checks (HANA_Configuration_MiniChecks*):

|M0890|Unusual frequent thread methods (last hour)    |        |IndirectScanBvOutJob<BV> (5.33 threads)                      |none          |X| 2114710|
|M1125|Columns with many scanned records              |        |MSEG.BUDAT_MKPF (41067249/s)                                 |none          |X| 2000002|
|M1125|Columns with many scanned records              |        |MSEG.MANDT (16265794/s)                                      |none          |X| 2000002|
|M1125|Columns with many scanned records              |        |MSEG.MATNR (375450570/s)                                     |none          |X| 2000002|
|M1125|Columns with many scanned records              |        |MSEG.MJAHR (354290653/s)                                     |none          |X| 2000002|
|M1125|Columns with many scanned records              |        |MSEG.WERKS (28137626/s)                                      |none          |X| 2000002|

Indication of necessity of an index

Thread methods like

◉ IndirectScanBvOutJob*
◉ JobParallelMgetSearch
◉ JobParallelPagedMgetSearch
◉ PrefixedScanVecOutJob
◉ PrefixedScanVecOutJob<range>
◉ RlePredScanJob<ScanVectorBinSearchPredicate>(out=vector)
◉ RlePredScanJob<ScanVectorPredicate>(out=vector)
◉ RleScanBvOutJob<BV>
◉ RleScanBvOutJob<range>
◉ RleScanVecOutJob<BV>
◉ RleScanVecOutJob<range>
◉ RleScanBvOutJob
◉ scanWithoutIndex
◉ ClusterIndexScanBvOutJob<ScanRangePredicate>
◉ ClusterScanBvOutJob<BV>
◉ ClusterScanBvOutJob<range>
◉ ClusterScanVecOutJob<range>
◉ SparseBvScanBvOutJob
◉ SparseBvScanVecOutJob
◉ SparsePredScanBvOutJob<ScanRangesPredicate>
◉ SparsePredScanVecOutJob<ScanRangesPredicate>
◉ SparsePredScanVecOutJob<ScanVectorBinSearchPredicate>
◉ SparsePredScanVecOutJob<ScanVectorPredicate>
◉ SparseRangeScanBvOutJob
◉ SparseRangeScanVecOutJob
◉ sparseSearch
◉ sse_icc_lib::mgetSearchi_AVX2impl
◉ sse_icc_lib::mgetSearchi_AVX

If you have found out the thread method, you have to go for the statement hash:

HANA_Threads_ThreadSamples_AggregationPerTimeSlice*

          ( SELECT                                                      /* Modification section */
              '1000/10/18 07:58:00' BEGIN_TIME,                  /* YYYY/MM/DD HH24:MI:SS timestamp, C, C-S<seconds>, C-M<minutes>, C-H<hours>, C-D<days>, C-W<weeks>, E-S<seconds>, E-M<minutes>, E-H<hours>, E-D<days>, E-W<weeks>, MIN */
              '9999/10/18 08:05:00' END_TIME,                    /* YYYY/MM/DD HH24:MI:SS timestamp, C, C-S<seconds>, C-M<minutes>, C-H<hours>, C-D<days>, C-W<weeks>, B+S<seconds>, B+M<minutes>, B+H<hours>, B+D<days>, B+W<weeks>, MAX */
              'SERVER' TIMEZONE,                              /* SERVER, UTC */
              '%' HOST,
              '%' PORT,
              -1  THREAD_ID,
              '%' THREAD_TYPE,                /* e.g. 'SqlExecutor', 'JobWorker' or 'MergedogMonitor' */
              '%' THREAD_STATE,               /* e.g. 'Running', 'Network Read' or 'Semaphore Wait' */
              'RleScanVecOutJob%' THREAD_METHOD,
              '%' THREAD_DETAIL,
              '%' STATEMENT_HASH,
              '%' ROOT_STATEMENT_HASH,
              '%' STATEMENT_ID,
              '%' STATEMENT_EXECUTION_ID,
              '%' DB_USER,
              '%' APP_NAME,
              '%' APP_USER,
              '%' APP_SOURCE,
              '%' LOCK_TYPE,
              '%' LOCK_NAME,
              -1  NUMA_NODE,
              '%' CLIENT_IP,
              -1  CLIENT_PID,
              -1  CONN_ID,
              -1  MIN_SAMPLES_TOTAL,
              60 TIME_SLICE_S,
              'X' EXCLUDE_SERVICE_THREAD_SAMPLER,
              'X' EXCLUDE_NEGATIVE_THREAD_IDS,
              'X' EXCLUDE_PHANTOM_THREADS,
              ' ' EXCLUDE_EMPTY_STATEMENT_IDS,
              ' ' EXCLUDE_EMPTY_STATEMENT_EXECUTION_IDS,
              'CURRENT' DATA_SOURCE,
              'HASH' AGGREGATE_BY,     /* SAMPLE_TIME, HOST, PORT, THREAD_ID, THREAD_TYPE, THREAD_METHOD, THREAD_DETAIL, THREAD_STATE, STATE_LOCK, HASH, ROOT, DB_USER, APP_NAME, APP_USER, APP_SOURCE, CLIENT_IP, CLIENT_PID, CONN_ID, LOCK_TYPE, LOCK_NAME, STATEMENT_ID, STAT_EXEC_ID, NUMA_NODE */
              -1 RESULT_ROWS

=> in this example thread method RleScanVecOutJob (may be you have to adjust the data source to HISTORY)

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

If you have found out the hash, you can translate it readable to SQL text via

HANA_SQL_StatementHash_SQLText:

  ( SELECT                         /* Modification section */
      '426bdc1658f5f21d5bebf85ec4ab9c2c' STATEMENT_HASH,
      60 LINE_LENGTH_TARGET,
      ' ' ADD_BLANK_AFTER_COMMA,
      ' ' REPLACE_BINDS,
      'EXPENSIVE_STATEMENTS' BIND_VALUES

426bdc1658f5f21d5bebf85ec4ab9c2c
SELECT COUNT(*) FROM "VBEP" WHERE "MANDT" = ? AND "VBELN" =
? AND "EDATU" = ?

From the ColStats you can see which col is frequently scanned and has no index (=>INDEX_TYPE=NONE):

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

At first have a look at the where clause. The ideal index would be on MANDT, VBELN and EDATU. But you should choose as less columns as possible. For this you can calculate the selectivity as stated out in this blog. Alternatively, you test more variation with the performance comparison (see below).

In most of the systems with only one client you can create a single col index on EDATU (due selectivity). In our case we have more than 20 clients which makes it worth to add the client column MANDT to the index.

Result in a multi columns index on:

◉ MANDT
◉ EDATU

6. Performance comparison


A before/after comparison is possible with the SQL: HANA_SQL_StatementHash_DiffReport

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

The difference of cursor and execution time is that the cursor time includes the client time.

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

6.1 Another example incl. thread method changes

Before

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

******************
* THREAD SAMPLES *
******************

SAMPLES PERCENT HOST                           PORT THREAD_TYPE
======= ======= ====================
   2144   68.30 hostname                      30040 JobWorker
    995   31.69 hostname                      30040 SqlExecutor

SAMPLES PERCENT THREAD_STATE                        LOCK_NAME                                                            
======= ======= ====================
   2618   83.40 Running                             ?
    518   16.50 Job Exec Waiting                    JobBarrier Wait for Jobs
      3    0.09 Job Exec Waiting                    ?

SAMPLES PERCENT THREAD_TYPE          THREAD_METHOD                                    
======= ======= ===================
   1988   63.33 JobWorker            SparseIndexScanVecOutJob
    995   31.69 SqlExecutor          ExecQidItab
     86    2.73 JobWorker            IndirectIndexScanVecOutJob
     65    2.07 JobWorker            RleScanVecOutJob
      5    0.15 JobWorker            

SAMPLES PERCENT THREAD_TYPE          THREAD_METHOD                                      THREAD_DETAIL                                                                  
======= ======= ===================
   1988   63.33 JobWorker            SparseIndexScanVecOutJob       ?
    995   31.69 SqlExecutor          ExecQidItab                                        ?
     86    2.73 JobWorker            IndirectIndexScanVecOutJob     ?
     65    2.07 JobWorker            RleScanVecOutJob                            ?
      5    0.15 JobWorker                                               

63% SparseIndexScanVecOutJob
32% ExecQidItab
3% IndirectIndexScanVecOutJob

Nearly 2/3 of the time were spend on thread methods which are an indicator for a need of an index.

After

SAP HANA Tutorial and Material, SAP HANA Learning, SAP HANA Exam Prep, SAP HANA Prep

Executions 15100
Records 474 0,03
Preparations 2 0
STAT_NAME    VALUE VALUE_PER_EXEC VALUE_PER_ROW
Elapsed time     0,34 h 82,8 ms           2637,73 ms
Execution time 0,34 h 82,77 ms           2636,92 ms
Preparation time 0     h   0,02 ms              0,81 ms
Lock wait time 0     h   0 ms                  0     ms


******************
* THREAD SAMPLES *
******************

SAMPLES PERCENT HOST                           PORT THREAD_TYPE
======= ======= ==================
     13  100.00 hostname                      30040 SqlExecutor

SAMPLES PERCENT THREAD_STATE                        LOCK_NAME                                                             
======= ======= ==================
     13  100.00 Running                             ?

SAMPLES PERCENT THREAD_TYPE          THREAD_METHOD                                    
======= ======= ==================
     13  100.00 SqlExecutor          ExecQidItab

SAMPLES PERCENT THREAD_TYPE          THREAD_METHOD                                      THREAD_DETAIL                                                                  
======= ======= =================
     13  100.00 SqlExecutor          ExecQidItab                                        ?
=> 100% ExecQidtab

=> but still room for improvement, but not for the execution plan => optimize compression

7. Tools


7.1 SHDB_INDEX_ANALYZE

The report SHDB_INDEX_ANALYZE from SAP note 1794297 should be executed as post conversion task.

“SHDB_INDEX_ANALYZE analyzes the data distribution of the largest tables in the system and suggests indexes on those fields which are highly selective and part of secondary indexes defined for traditional data bases. This report analyzes SAP and customer tables and indexes. This report has to run on the system already ported to HANA and should contain the full production data.”

7.2 SHDB_INDEX_CREATE

“SHDB_INDEX_CREATE creates secondary indexes based on a suggestion list imported into the select option INDFLD. The naming convention of this list per entry is
-Three characters to identify the SAP or customer namespace (SAP or. CUS)
-table name
-“~”
-field name
If available this report should run in the corresponding development system. The created indexes will then be transported through the landscape with standard mechanisms.”

7.3 SHDB_INDEX_CHECK

SHDB_INDEX_CHECK checks the created Indexes and ensures that all required indexes are created on the HANA data base. Additionally, the corresponding transport request entries are created.

7.4 indexAdvisor.py

Use the index adviser to find out for which tables and columns indexing would be most valuable. The indexAdvisor.py script is part of a SAP HANA system installation and runs from the command line. It is located in the $DIR_INSTANCE/exe/python_support directory.

7.5 SQL Collection

The best tool to analyze the performance and determine the col combinations for an index are statements in note 1969700. Get familiar with them and combine it with the know-how of the most popular FAQ HANA notes. There is no silver bullet as general recommendation.

No comments:

Post a Comment