PROFILE

Syntax

SQL

PROFILE <statement>;

Note

See the Remarks below for information about supported statements.

SQL

SHOW PROFILE [<JSON>] [PLAN <plan_id>] [PROCESS <process_id>] [INTO OUTFILE <file_name>] [UI] [ON NODE <node_id>];

Remarks

Prior to the 8.1 release, PROFILE could only be utilized on a manual basis. The command was appended to the beginning of the query to be profiled as shown above. Once the query finished executing, the SHOW PROFILE command was run to obtain the profile results. This profiling method is called an explicit profile since the PROFILE command is explicitly written with the query.
Now SingleStore offers two automatic profiling (called "auto profiling" hereafter) methods: FULL and LITE. Auto profiling and explicit profiling share the same query plans, but the difference between them is the level of statistic collecting. Currently, auto profiling is not the default setting. See Query Performance Tools for more details about auto profiling.

INSERT...SELECT, REPLACE...SELECT query shapes work with auto profiling. Aggregator result tables cannot be auto profiled.
When you run a PROFILE statement, it executes the associated statement and collects resource usage metrics about the statement. The PROFILE command works with SELECT, UPDATE, INSERT ... SELECT, and DELETE statements, but not INSERT statements.
After the query has executed, run the SHOW PROFILE statement to display the collected metrics. Like the EXPLAIN statement, SHOW PROFILE displays query execution operations as a tree, where the operator at the top is the last executed before returning the result to the client, and operators below are executed before their parents; however, SHOW PROFILE additionally displays resource usage metrics for the operators in the execution tree.
To get more accurate resource usage metrics of a query, run the PROFILE statement twice followed by the SHOW PROFILE statement.

Note

You must connected to the master aggregator when running this command on reference tables.

The SHOW PROFILE JSON command returns the profile information of a query in JSON (machine readable format). To get this information, run the PROFILE command followed by the SHOW PROFILE JSON command. This information includes detailed statistics about some operators and statistics on compilation time (displayed under compile_time_stats). See Example1 for details.
If the variable profile_for_debug is set to ON, SHOW PROFILE JSON will output additional data that is useful for troubleshooting query optimizer issues. See more about this variable in the Non-Sync Variables List. See all possible additional data output in PROFILE JSON Metrics link. See an example of this below as well.
Run the SHOW PROFILE PLAN command with a specific plan_id to display profile information (optionally in JSON format) for any plan in the cache that was run in PROFILE mode. For information related to plan_id, refer to the SHOW PLAN topic.
SQL
```
---To display profile information for a specific plan
SHOW PROFILE PLAN 4;
```
SQL
```
---To display the profile information for a specific plan in JSON format
SHOW PROFILE JSON PLAN 4;
```

To write the result of a profile query to a file, use the INTO OUTFILE file_name clause. It creates a new file at the specified location and writes the query profile details to the file. Enter the absolute path to the file as the file_name parameter. The output file can be a JSON or a text file.

Note

If the file already exists, then SingleStore will throw an error.

To write the profile information of a query in JSON format to a file:
SQL
```
SHOW PROFILE JSON INTO OUTFILE '/tmp/testprof.json';
```

To write the profile information of a query to a text file:
```
SHOW PROFILE INTO OUTFILE '/tmp/testprof.txt';
```

To write the profile statistics of a specific plan ID to a file:
SQL
```
SHOW PROFILE PLAN 7 INTO OUTFILE '/tmp/profile_plan7.txt';
```

To write the profile statistics of a specific plan ID in JSON format to a file:
SQL
```
SHOW PROFILE JSON PLAN 7 INTO OUTFILE '/tmp/profjson_plan7.json';
```

The SHOW PLAN [JSON] plan_id command displays the EXPLAIN plan of a query. Refer to SHOW PLAN for details.

Profiling a Hung Query

To get information on progress of a query and to debug queries with long execution times, run the PROFILE statement in one connection and SHOW PROFILE PROCESS in another connection. For example, run the following statements in one connection:

SQL

PROFILE select_statement;

SQL

SHOW PROFILE;

Simultaneously, open another connection and run the SHOW PROFILE PROCESS statement while the PROFILE statement is in progress:

SQL

SHOW PROFILE PROCESS process_id;

You can get the process ID from the SHOW PROCESSLIST command or the CONNECTION_ID function as follows:

SQL

SHOW PROCESSLIST;

SQL

SELECT CONNECTION_ID();

If there is a skew between different partitions, then SHOW PROFILE displays additional details of the skewed metric. A skewed partition displays in the output of a SHOW PROFILE query in the following format: [memory_usage: v | max:w at partition_x, average: y, std dev: z] where,

The total across all partitions is v.
The partition with the largest amount is partition x with memory use w.
The average memory usage per partition is y.
The standard deviation is z.

If all data is on one partition, then the actual_rows has the same count across all partitions as max on a single partition. The output of SHOW PROFILE JSON always displays these additional details, even when there is no skew.

`PROFILE` Metrics

The following table provides a brief explanation of the metrics that are gathered when executing a statement using PROFILE.

Metric	Description
exec_time	Time spent in running each operator. For HashJoin, exec_time is calculated as the time spent in probing the hash table. `build_exec_time` (specific to HashJoin operator) shows the time spent in building the hash table
network_time	Wait time spent while flushing data to the network
memory_usage	Memory used by the operator in KB
network_traffic	Data sent over the network in KB
actual_rows	Number of rows processed by the operator. For example, for IndexRangeScan, actual_rows is the number of rows scanned
est_table_rows	Estimated number of rows in a table. This is an attribute of a table, not of an operator
est_filtered	Estimated number of rows of a table after applying all single-table filters
start_time	Time difference between the query start and operator execution as start_time:hh:mm:ss.SSS. Alternatively, in start_time: [hh:mm:ss.SSS, hh:mm:ss.SSS] format, it shows maximum and minimum start time over all threads (partitions)
end_time	Time when the operator finishes execution

Not all of the above metrics will be gathered for queries where execution performance may be hindered by gathering such metrics.

`PROFILE` Metrics for Columnstore Queries that Use Filters

The following table provides a brief explanation of the columnstore metrics that are gathered when running PROFILE with a columnstore query that uses filters. These metrics appear after running SHOW PROFILE JSON, following PROFILE.

Metric	Description
columnstore_filter_execution_type	The type of filter used (full row scan, hash index, or bloom).
columnstore_filter_total_rows_in	The number of rows that the filter predicate was evaluated on.
columnstore_filter_total_rows_out	The number of rows that passed the filter predicate.
columnstore_filter_encoded_rows_in	The number of rows that the filter predicate was evaluated on, using operations on encoded data.
columnstore_filter_encoded_rows_out	The number of rows that passed the filter predicate, using operations on encoded data.
columnstore_filter_avg_filters_per_row	The number of average filtering operations performed on each row. SingleStore combines filtering options where possible, but outside of these cases, the following logic is true: if two columns each had a filter applied to them on every row, the `columnstore_filter_avg_per_row` would be 2.0.
columnstore_filter_avg_index_filters_per_row	This is the same as `columnstore_filter_avg_filters_per_row`, but instead looking at filtering on indices, rather than all filters as a group.
columnstore_filter_avg_bloom_filters_per_row	This is the same as `columnstore_filter_avg_filters_per_row`, but instead looking at how many rows are filtered out through use of a bloom filter, rather than all filters as a group.

Whether or not a filter predicate was evaluated using operations on encoded data depends on the filter predicate itself as well as the encodings of the involved columns. If columnstore_filter_encoded_rows_in is lower than columnstore_filter_total_rows_in, either the filter predicate does not support operations on encoded data or there were some segments with encodings that do not support operations on encoded data.

Some filter predicates are children of other filter predicates. This is signaled by indentation levels in the text of the condition attribute. During filter evaluation, either the parent filter predicate is used, or its children are evaluated instead. The columnstore_filter_total_rows_in metric can be inspected to find which method of evaluation is used in what proportion.

Important

Up-to-date PROFILE metrics for columnstore filters are available after the columnstore background flusher has run. The flusher runs automatically on regular intervals. You can also run it manually via OPTIMIZE TABLE ... FLUSH.

See an example that provides resource usage metrics for a columnstore query that uses filters.

`PROFILE` Blob Cache Metrics

For columnstore tables stored in unlimited storage, there are additional metrics to provide information about the blob cache.

Metric	Description
blobcache_total_blobs_downloaded	The number of blobs downloaded from unlimited storage due to local blob cache misses. The metric sums up the number of downloaded blobs from all leaf nodes. The metric has the following sub-categories: blobcache_columnar_blobs_downloaded blobcache_inverted_index_blobs_downloaded blobcache_cross_segment_hash_index_blobs_downloaded blobcache_fts_blobs_downloaded blobcache_column_group_blobs_downloaded
blobcache_total_blobs_downloaded_bytes	The number of blob bytes downloaded from unlimited storage due to local blob cache misses. The metric sums up the downloaded blob bytes from all leaf nodes. The metric has the following sub-categories: blobcache_columnar_blobs_downloaded_bytes blobcache_inverted_index_blobs_downloaded_bytes blobcache_cross_segment_hash_index_blobs_downloaded_bytes blobcache_fts_blobs_downloaded_bytes blobcache_column_group_blobs_downloaded_bytes
blobcache_total_blobs_accessed	The number of blobs accessed by the query. A blob is "accessed" when the query reads some data from it, regardless of how much data there is. The metric sums up the number of accessed blobs from all leaf nodes. The metric includes blob files that were both initially in the blob cache (cache hit) and downloaded from unlimited storage (cache miss). "Accessed” means the query has read some data from the blob. The metric has the following sub-categories: blobcache_columnar_blobs_accessed blobcache_inverted_index_blobs_accessed blobcache_cross_segment_hash_index_blobs_accessed blobcache_fts_blobs_accessed blobcache_column_group_blobs_accessed
blobcache_total_blobs_accessed_bytes	The number of blob bytes accessed by the query. The metric sums up the number of accessed blob bytes from all leaf nodes. The metric has the following sub-categories: blobcache_columnar_blobs_accessed_bytes blobcache_inverted_index_blobs_accessed_bytes blobcache_cross_segment_hash_index_blobs_accessed_bytes blobcache_fts_blobs_accessed_bytes blobcache_column_group_blobs_accessed_bytes
blobcache_wait_time_ms	The time that was spent on waiting for blob cache data. The metric is the maximum waiting time (in milliseconds) from all query threads. To be more specific, query threads run concurrently with blob-downloading threads. This metric measures the time that query threads spend waiting for blob cache data. This metric does not measure the time that download worker threads spend downloading from the network.

`PROFILE JSON` Fields

When unlimited is enabled, PROFILE JSON provides information about the segments and blobs retrieved from unlimited storage to the blob cache (on local disk) under the ColumnStoreScan executor. This information includes segment statistics covering timing and size information of blobs.

Field	Description
segments_scanned	The number of segments that were scanned.
segments_skipped	The number of segments that were not scanned due to segment elimination.
segments_fully_contained	The number of segments that were not scanned, because they are fully contained.
segments_in_blob_cache	Represents the number of segments in the blob cache, which did not need retrieval from unlimited storage.
number_of_blocks_tested_for_block_elim	The number of blocks on which sub-segment elimination was run.
number_of_blocks_eliminated_for_block_elim	The number of blocks that were eliminated by sub-segment elimination.
column_group_project_rows	The number of rows scanned using the row index (COLUMN GROUP).
column_segment_project_rows	The number of rows scanned when scanning the individual column segments.
blob_fetch_network_time	The total time spent waiting for blobs to be retrieved over the network.
total_blobs_fetched_size	The total size of blobs retrieved from unlimited storage and stored in the blob cache.
blobs_fetched_stats	Provides detail about various blob sizes and types retrieved. Includes the following fields: columnar_blobs_fetched_size - provides the size of the columnar blobs retrieved. inverted_index_blobs_fetched_size - provides the size of the index blobs within segments. cross_segment_hash_index_blobs_fetched_size - provides the size of index blobs across segments. fts_blobs_fetched_size - provides the size of the full- text search index blobs. column_group_xxxx - provides the size of the row index (COLUMN GROUP) blobs retrieved.

`PROFILE JSON` Metrics with `profile_for_debug` = `ON`

The following list shows all additional data included in the output of SHOW PROFILE JSON when it is run with the variable set_profile_for_debug set to ON. Different data is displayed, depending on what is relevant to the query being profiled. For example, in the case that tables included in a query do not have any statistics collected on them, the table statistics metadata will be empty in the output. The debugging information consists of:

DDL for all database objects relevant to the query
Global variable settings
Session variable settings
Table statistics metadata
Sampling cache for filter selectivities
Sampling cache for row counts
Autostats results cache

The following commands can be used without needing to set the profile_for_debug session variable.

SQL

PROFILE REPRO <query>;
SHOW REPRODUCTION;

SingleStore will continue to support the profile_for_debug session variable with the PROFILE and SHOW PROFILE JSON commands for backward compatibility.

Note that the profile_for_debug session variable can only be used on fully or partially executed queries. To learn how to collect troubleshooting data for queries that fail in the compilation phase, refer to the SHOW REPRODUCTION command reference.

`PROFILE JSON` Metrics with `enable_spilling` = `ON`

Sometimes, a GROUP BY query will use a hash table internally to execute the query, which is known as a HashGroupBy operation. For this operation, when node memory load is above 75% and the hash table (for the current thread) is at least 100MB, the intermediate result of the operation spills to disk in the data/spill directory. Spilling will usually reduce memory usage (can spill at most 64x physical memory), but may result in longer execution time.

Note: SingleStore recommends to use an SSD for the spilling directory, since it can result in a tenfold increase in performance.

The output of SHOW PROFILE JSON also includes the following metrics, when the sync variable enable_spilling is set to ON (starting with version 7.8, ON is the default value for this sync variable):

spill_outputted_rows: The number of rows spilled to disk
spill_disk_usage : The number of bytes spilled to disk

The sync variables spilling_node_memory_threshold_ratio , spilling_query_operator_memory_threshold , and spilling_minimal_disk_space (starting with version 7.8) also define the threshold limit to start spilling. Refer to Sync Variables Lists for details.

Visual Profile via SHOW PROFILE UI

Using SHOW PROFILE UI will generate a URL that loads a visual representation of the query profile. When you open the link, the Visual Explain site is loaded and the query profile is displayed.

Examples

Example 1

The SHOW PROFILE JSON statement displays the compilation and optimizer statistics. This helps to troubleshoot slow query performance and compare the time spent in running versus compiling the query. In this example, the variable profile_for_debug is set to OFF.

For example, run the following PROFILE query:

SQL

PROFILE SELECT * from Emp;

+-----------+---------+
| Name      | City    |
+-----------+---------+
| Sam       | Chicago |
| Jack      | Norway  |
| Tom       | Chicago |
| Neil      | Chicago |
+-----------+---------+
4 rows in set (0.29 sec)

Now run the SHOW PROFILE JSON command. The following output is filtered to display only the compilation and optimization details. It has also been tab spaced to increase readability, and may differ from the actual output.

SQL

SHOW PROFILE JSON;

+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
"query_info":{
       "query_text":"profile select * from Emp",
       "total_runtime_ms":"291",
       "text_profile":"Gather partitions:all alias:remote_0 actual_rows: 9 exec_time: 0ms start_time: 00:00:00.001 end_time: 00:00:00.143\nProject [Emp.Name, Emp.City] actual_rows: 9 exec_time: 0ms start_time: 00:00:00.139 network_traffic: 0.100000 KB network_time: 1ms\nTableScan trades.Emp actual_rows: 9 exec_time: 0ms start_time: 00:00:00.139\nCompile Total Time: 147ms\n",
       "compile_time_stats":{
           "mbc_emission":"0",
           "create_mbc_context":"4",
           "expanding_views":"0",
           "optimizer_query_rewrites":"1",
           "optimizer_stats_analyze":"0",
           "optimizer_stats_other":"0",
           "optimizer_stats_sampling":"0",
           "optimizer_setting_up_subselect":"0",
           "optimizer_distributed_optimizations":"0",
           "optimizer_enumerate_temporary_tables":"0",
           "optimizer_singlebox_optimizations_agg":"0",
           "optimizer_rewrite_estimate_cost":"0",
           "optimizer_stats_autostats":"0",
           "ppc_check":"0",
           "generating_query_mpl":"1",
           "generating_user_function_mpl":"0",
           "module_cleaning":"0",
           "dump_clean_bitcode":"0",
           "llvm_optimization":"0",
           "dump_optimized_bitcode":"0",
           "machine_code_generation":"0",
           "llvm_printing":"0",
           "asm_printing":"0",
           "symbol_resolution":"0",
           "unknown":"138",
           "total":"147"
       }
   }
} |
+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+-
1 row in set (0.00 sec)

If you had run PROFILE again prior to running SHOW PROFILE JSON, the results would have been more accurate.

Note: The PROFILE command can be used for long running queries to analyze the query plan and search for unexpectedly expensive operators. Output the result in JSON format using the SHOW PROFILE JSON command to analyze the query plan visually in the Visual Explain UI.

Example 2

This example shows how to run SHOW PROFILE JSON with the variable profile_for_debug set to ON in order to get additional data output beyond what is displayed in the first example where this variable is set to OFF. Different data is displayed, depending on what is relevant to the query being profiled.

SQL

SET profile_for_debug = on;
PROFILE SELECT * FROM t1 JOIN t2 ON t1.b = t2.b;
SHOW PROFILE JSON;

/* The following JSON output has been truncated to show only the additional data provided with set_profile_for_debug set to ON. */
"debug_info": {
    "ddl": [
      "CREATE DATABASE `db` PARTITIONS 3",
      "USING `db` CREATE TABLE `t2` (  `a` bigint(11) NOT NULL AUTO_INCREMENT,  `b` int(11) DEFAULT NULL,  PRIMARY KEY (`a`)) AUTOSTATS_CARDINALITY_MODE=OFF AUTOSTATS_HISTOGRAM_MODE=OFF SQL_MODE=\'STRICT_ALL_TABLES\'",
      "USING `db` CREATE TABLE `t1` (  `a` bigint(11) NOT NULL AUTO_INCREMENT,  `b` int(11) DEFAULT NULL,  PRIMARY KEY (`a`)) AUTOSTATS_CARDINALITY_MODE=OFF AUTOSTATS_HISTOGRAM_MODE=OFF SQL_MODE=\'STRICT_ALL_TABLES\'"
    ],
    "optimizer_stats": [],
    "sampling_rowcount_cache": [
      {
        "db": "db",
        "rowcount": "1",
        "table": "t1"
      },
      {
        "db": "db",
        "rowcount": "4",
        "table": "t2"
      }
    ],
    "variables": {
      "as_aggregator": 1,
      "cardinality_estimation_level": 3,
      "client_found_rows": 0,
      "collation_server": 2,
      "comment from t.py": "t.py normalize_profile_json() is modifying this output - most variables are filtered out of this output, look at t.py for details",
      "data_conversion_compatibility_level": 0,
      "default_partitions_per_leaf": 3,
      "disable_sampling_estimation_with_histograms": 2,
      "disable_subquery_merge_with_straight_joins": 2,
      "distributed_optimizer_min_join_size_run_initial_heuristics": 18,
      "force_bushy_joins": 0,
      "interpreter_mode": 4,
      "json_extract_string_collation": 3,
      "materialize_ctes": 0,
      "max_subselect_aggregator_rowcount": 0,
      "resource_pool_is_auto": 0,
      "resource_pool_statement_selector_function": 4294967300,
      "sql_mode": 4194304,
      "sql_select_limit": 1
    }
  },

This example shows how to run PROFILE REPRO followed by SHOW REPRODUCTION commands that will replace the SET profile_for_debug=1; PROFILE <query>; and SHOW PROFILE JSON series of commands.

SQL

PROFILE REPRO SELECT * FROM t; 
SHOW REPRODUCTION;

REPRODUCTION INFO
/* The following JSON output has been truncated to show only the additional data provided with set_profile_for_debug set to ON. */
    "debug_info":{
        "ddl":[
    "CREATE DATABASE `test1` PARTITIONS 2",
    "USING `test1` CREATE TABLE `t` (\n  `id` bigint(11) NOT NULL AUTO_INCREMENT,\n  `a` int(11) DEFAULT NULL,\n  `b` int(11) DEFAULT NULL,\n  UNIQUE KEY `PRIMARY` (`id`) USING HASH,\n  SHARD KEY `__SHARDKEY` (`id`),\n  KEY `__UNORDERED` () USING CLUSTERED COLUMNSTORE\n) AUTO_INCREMENT=1000000 AUTOSTATS_CARDINALITY_MODE=INCREMENTAL AUTOSTATS_HISTOGRAM_MODE=CREATE AUTOSTATS_SAMPLING=ON SQL_MODE='STRICT_ALL_TABLES'"
],
        "variables":{"as_aggregator": 1, "as_leaf": 0, "batch_external_functions": 0, "binary_serialization": 1, "client_found_rows": 1, "collation_server": 2, "datetime_precision_mode": 0, "default_columnstore_table_lock_threshold": 0, "disable_histogram_estimation": 0, "disable_sampling_estimation": 0, "disable_sampling_estimation_with_histograms": 2, "disable_subquery_merge_with_straight_joins": 2, "display_full_estimation_stats": 0, "distributed_optimizer_broadcast_mult": 0, "distributed_optimizer_estimated_restricted_search_cost_bound": 125, "distributed_optimizer_max_join_size": 22, "distributed_optimizer_min_join_size_run_initial_heuristics": 16, "distributed_optimizer_nodes": 0, "distributed_optimizer_old_selectivity_table_threshold": 22, "distributed_optimizer_run_legacy_heuristic": 0, "distributed_optimizer_selectivity_fallback_threshold": 50000000, "distributed_optimizer_unrestricted_search_threshold": 22, "enable_broadcast_left_join": 1, "enable_local_shuffle_group_by": 1, "enable_multipartition_queries": 1, "enable_skiplist_sampling_for_selectivity": 1, "force_bloom_filters": 0, "force_bushy_join_table_limit": 18, "force_bushy_joins": 0, "force_heuristic_rewrites": 0, "force_table_pushdown": 0, "hash_groupby_segment_distinct_values_threshold": 10000, "ignore_insert_into_computed_column": 0, "inlist_precision_limit": 10000, "interpreter_mode": 4, "leaf_pushdown_default": 0, "leaf_pushdown_enable_rowcount": 120000, "materialize_ctes": 1, "max_broadcast_tree_rowcount": 120000, "max_subselect_aggregator_rowcount": 120000, "old_local_join_optimizer": 0, "optimize_constants": 1, "optimize_mpl_before_printing": 0, "optimize_stmt_threshold": 50, "optimizer_beam_width": 10, "optimizer_cross_join_cost": 1.000000, "optimizer_disable_right_join": 0, "optimizer_disable_subselect_to_join": 0, "optimizer_empty_tables_limit": 0, "optimizer_hash_join_cost": 1.000000, "optimizer_merge_join_cost": 1.000000, "optimizer_nested_join_cost": 1.000000, "optimizer_num_partitions": 0, "quadratic_rewrite_size_limit": 200, "query_rewrite_loop_iterations": 1, "reshuffle_group_by_base_cost": 0, "sampling_estimates_for_complex_filters": 1, "singlebox_optimizer_cost_based_threshold": 18, "sql_mode": 4194304, "sql_select_limit": 0, "subquery_merge_with_outer_joins": 3, "cardinality_estimation_level": 3, "data_conversion_compatibility_level": 0, "debug_mode": 0, "default_partitions_per_leaf": 1, "disable_update_delete_distributed_transactions": 0, "enable_alias_space_trim": 0, "enable_spilling": 1, "explicit_defaults_for_timestamp": 1, "json_extract_string_collation": 3, "resource_pool_statement_selector_function": 4294967300, "use_avx2": 1, "use_dstree": 1, "use_joincolumnstore": 1, "node_degree_of_parallelism": 0, "resource_pool_is_auto": 0},
        "optimizer_stats":[
    {
    "version": 2,
    "databaseName": "test1",
    "tables": [
        {
            "tableName": "t",
            "rowCount": "2048",
            "columns": [
                {
                    "columnName": "id",
                    "nullCount": "0",
                    "minValue": "",
                    "maxValue": "",
                    "cardinality": "2048",
                    "density": "0x0p+0",
                    "sampleSize": "0",
                    "lastUpdated": "1646329049"
                },
                {
                    "columnName": "a",
                    "nullCount": "0",
                    "minValue": "",
                    "maxValue": "",
                    "cardinality": "101",
                    "density": "0x0p+0",
                    "sampleSize": "0",
                    "lastUpdated": "1646329049"
                },
                {
                    "columnName": "b",
                    "nullCount": "0",
                    "minValue": "",
                    "maxValue": "",
                    "cardinality": "101",
                    "density": "0x0p+0",
                    "sampleSize": "0",
                    "lastUpdated": "1646329049"
                }
            ]
        }
    ],
    "stats": [
        {
            "tableName": "t",
            "columns": [
                {
                    "columnName": "id"
                },
                {
                    "columnName": "a"
                },
                {
                    "columnName": "b"
                }
            ]
        }
    ]
}
],
        "query_after_rewrites":"SELECT `t`.`id` AS `id`, `t`.`a` AS `a`, `t`.`b` AS `b` FROM  `test1`.`t` as `t`  LIMIT @@SESSION.`sql_select_limit` \/*!90623 OPTION(CLIENT_FOUND_ROWS=1)*\/"
        }
    }

Example 3

The INTO OUTFILE file_name clause is useful to output the profile information of a query/plan to a file. In this example, the variable profile_for_debug is set to OFF.

For example, run the following PROFILE query:

SQL

PROFILE SELECT * from Emp;

Now, run the following command to write the JSON profile information to the EmpProfile.json file in the /tmp directory.

SQL

SHOW PROFILE JSON INTO OUTFILE '/tmp/EmpProfile.json';

To view the contents of the EmpProfile.json file, run the following command at the root prompt:

cat '/tmp/EmpProfile.json';

Example 4

The PROFILE statement is particularly helpful when evaluating distributed query performance, such as a distributed join. Distributed joins that require broadcasts or repartitions are expensive, and the PROFILE statement can help you understand how such queries are executed so that they can be optimized. In this example, the variable profile_for_debug is set to OFF.

In the following example, a distributed join is executed to return data about customers and their orders. For demonstration purposes, we loaded simulated statistics for each table.

Suppose we have run the query:

SQL

SELECT COUNT(*)
FROM orders o JOIN customer c
WHERE o.custkey = c.custkey and c.mktsegment = 'BUILDING';

Now we want to diagnose why it is slower than expected. It’s best to first execute an EXPLAIN statement to understand the execution plan for a query.

SQL

EXPLAIN SELECT COUNT(*) FROM orders o JOIN customer c WHERE o.custkey = c.custkey and c.mktsegment = 'BUILDING';

+------------------------------------------------------------------------------------------------------------------------------------------------+
| EXPLAIN                                                                                                                                        |
+------------------------------------------------------------------------------------------------------------------------------------------------+
| Project [CAST(COALESCE($0,0) AS SIGNED) AS `count(*)`]                                                                                         |
| Aggregate [SUM(remote_0.`count(*)`) AS $0]                                                                                                     |
| Gather partitions:all est_rows:1 alias:remote_0                                                                                                |
| Project [`count(*)`] est_rows:1 est_select_cost:87,206,400                                                                                     |
| Aggregate [COUNT(*) AS `count(*)`]                                                                                                             |
| NestedLoopJoin                                                                                                                                 |
| |---IndexRangeScan orders AS o, KEY orders_fk1 (o_custkey) scan:[o_custkey = r1.c_custkey] est_table_rows:150,000,000 est_filtered:150,000,000 |
| TableScan r1 storage:list stream:no est_table_rows:2,725,200                                                                                   |
| Broadcast [c.c_custkey] AS r1 distribution:direct est_rows:2,725,200                                                                           |
| Filter [c.c_mktsegment = ?]                                                                                                                    |
| TableScan customer AS c, PRIMARY KEY (c_custkey) est_table_rows:150,000,000 est_filtered:2,725,200                                             |
+------------------------------------------------------------------------------------------------------------------------------------------------+

From this EXPLAIN statement, it’s clear that a broadcast is required, and that a nested loop join will be performed. Now you can run the PROFILE statement with the same SELECT query to gather resource usage metrics.

SQL

PROFILE SELECT COUNT(*) FROM orders o JOIN customer c WHERE o.custkey = c.custkey and c.mktsegment = 'BUILDING';

+----------+
| count(*) |
+----------+
| 421      |
+----------+

The PROFILE statement will output the same results as the inner SELECT statement, but it has also gathered resource usage metrics. To see the metrics, you must run the SHOW PROFILE statement.

SQL

SHOW PROFILE;

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| PROFILE                                                                                                                                                                                                                   |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Project [CAST(COALESCE($0,0) AS SIGNED) AS `count(*)`] actual_rows: 1 exec_time: 0ms start_time: 00:00:00.584 network_traffic: 0.004000 KB                                                                                |
| Aggregate [SUM(remote_0.`count(*)`) AS $0] actual_rows: 3 exec_time: 0ms start_time: 00:00:00.581                                                                                                                         |
| Gather partitions:all est_rows:1 alias:remote_0 actual_rows: 3 exec_time: 0ms start_time: 00:00:00.003 end_time: 00:00:00.583                                                                                             |
| Project [`count(*)`] est_rows:1 est_select_cost:87,206,400 actual_rows: 3 exec_time: 0ms start_time: [00:00:00.581, 00:00:00.582] network_traffic: 0.012000 KB                                                            |
| Aggregate [COUNT(*) AS `count(*)`] actual_rows: 421 exec_time: 0ms start_time: [00:00:00.053, 00:00:00.056]                                                                                                               |
| NestedLoopJoin actual_rows: 421 exec_time: 0ms                                                                                                                                                                            |
| |---IndexRangeScan orders AS o, KEY orders_fk1 (o_custkey) scan:[o_custkey = r1.c_custkey] est_table_rows:150,000,000 est_filtered:150,000,000 actual_rows: 421 exec_time: 532ms start_time: [00:00:00.053, 00:00:00.059] |
| TableScan r1 storage:list stream:no est_table_rows:2,725,200 actual_rows: 122,424 exec_time: 29ms start_time: [00:00:00.004, 00:00:00.005]                                                                                |
| Broadcast [c.c_custkey] AS r1 distribution:direct est_rows:27,252 actual_rows: 40,808 exec_time: 19ms start_time: 00:00:00.002 network_traffic: 334.326996 KB                                                             |
| Filter [c.c_mktsegment = ?] actual_rows: 40,808 exec_time: 97ms start_time: 00:00:00.002                                                                                                                                  |
| TableScan customer AS c, PRIMARY KEY (c_custkey) est_table_rows:150,000 est_filtered:27,252 actual_rows: 200,000 exec_time: 68ms start_time: 00:00:00.002                                                                 |
| Compile Total Time: 238ms
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

After the SHOW PROFILE statement has been executed, you can see from the example above that actual rows, network traffic, and execution time have been appended to the end of the execution operator lines. We can first identify that most expensive operator is IndexRangeScan orders, which takes 532ms. Then we can see that TableScan tpch.orders has many fewer rows now than are recorded in statistics, which suggests that our statistics are not up-to-date.

Similarly, we can identify that the most network-consuming operator is Broadcast [c.custkey] and there is no memory-consuming operator (every operator uses only a constant amount of memory).

After running ANALYZE TABLE, we can rerun the PROFILE query:

SQL

PROFILE SELECT COUNT(*) FROM orders o JOIN customer c where o.custkey = c.custkey and c.mktsegment = 'BUILDING';

+----------+
| count(*) |
+----------+
| 421      |
+----------+

Finally, we can output the profile information for it:

SQL

SHOW PROFILE;

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| PROFILE                                                                                                                                                                   |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Project [CAST(COALESCE($0,0) AS SIGNED) AS `count(*)`] actual_rows: 1 exec_time: 0ms start_time: 00:00:00.237 network_traffic: 0.004000 KB                                |
| Aggregate [SUM(remote_0.`count(*)`) AS $0] actual_rows: 3 exec_time: 0ms start_time: 00:00:00.223                                                                         |
| Gather partitions:all est_rows:1 alias:remote_0 actual_rows: 3 exec_time: 0ms start_time: 00:00:00.004 end_time: 00:00:00.236                                             |
| Project [`count(*)`] est_rows:1 est_select_cost:25,006 actual_rows: 3 exec_time: 0ms start_time: [00:00:00.221, 00:00:00.235] network_traffic: 0.012000 KB                |
| Aggregate [COUNT(*) AS `count(*)`] actual_rows: 421 exec_time: 0ms start_time: [00:00:00.140, 00:00:00.141]                                                               |
| HashJoin [r1.o_custkey = customer.c_custkey] actual_rows: 421 exec_time: 11ms start_time: [00:00:00.140, 00:00:00.141] memory_usage: 655.359985 KB                        |
| |---Broadcast [orders.o_custkey] AS r1 distribution:direct est_rows:12,503 actual_rows: 12,503 exec_time: 11ms start_time: 00:00:00.002 network_traffic: 78.333000 KB     |
| |   TableScan tpch.orders, PRIMARY KEY (o_orderkey) est_table_rows:12,503 est_filtered:12,503 actual_rows: 12,503 exec_time: 13ms start_time: 00:00:00.002                |
| Filter [customer.c_mktsegment = ?] actual_rows: 40,808 exec_time: 52ms start_time: [00:00:00.140, 00:00:00.141]                                                           |
| TableScan tpch.customer, PRIMARY KEY (c_custkey) est_table_rows:200,000 est_filtered:40,832 actual_rows: 200,000 exec_time: 41ms start_time: [00:00:00.140, 00:00:00.141] |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

The new plan is faster in execution time but notice that we introduced a memory-consuming operator HashJoin, which allocates 655.359985 KB of memory at run time in the example. That could be something to watch out for if we add more data, since it might limit the total number of queries that can run concurrently.

SQL

PROFILE SELECT name FROM company WHERE name LIKE "C%";

+------------------------------+
| name                         |
+------------------------------+
| Cutera, Inc.                 |
| CVB Financial Corporation    |
| CVD Equipment Corporation    |
| Cyanotech Corporation        |
| ...                          |
| Curis, Inc.                  |
+------------------------------+
348 rows in set (404 ms)

Now run the SHOW PROFILE JSON query to view the resource usage metrics in JSON format:

SQL

SHOW PROFILE JSON;

+------------------------------+
| PROFILE                      |
+------------------------------+
| {
    "profile":[
    {
            "executor":"Gather",
            "partitions":"all",
            "query":"SELECT `company`.`name` AS `name` FROM `trades_0`.`company`
             as `company`  WHERE (`company`.`name` LIKE 'C%')
             OPTION(NO_QUERY_REWRITE=1, INTERPRETER_MODE=LLVM)",
            "alias":"remote_0",
            "actual_row_count":{ "value":348, "avg":0.000000, "stddev":0.000000,
            "max":0, "maxPartition":0 },
            "actual_total_time":{ "value":0 },
            "start_time":{ "value":0 },
            "end_time":{ "value":5 },
            ...
            }
}
1 row in set (290 ms)

The SHOW PLAN command displays the plan information for the specified plan ID:

SQL

SHOW PLAN 5;

+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| PLAN                                                                                                                                                                    |
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Gather partitions:all alias:remote_0                                                                                                                                    |
| Project [company.symbol, company.name, company.last_sale, company.market_cap, company.IPO_year, company.sector, company.industry, company.summary_quote, company.extra] |
| TableScan trades.company                                                                                                                                                |
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
3 rows in set (0.00 sec)

Next, get the JSON plan information by running the following command:

SQL

SHOW PLAN JSON 5;

This command produces a large amount of text as result.

Example 5

The following example provides resource usage metrics for columnstore query that uses filters. In this example, the variable profile_for_debug is set to OFF.

Run this statement:

SQL

PROFILE SELECT * FROM products WHERE Color = 'Blue' and Qty = '5';

Then run SHOW PROFILE JSON.

Selected output:

JSON

...
"inputs":[
  {
    "executor":"ColumnStoreFilter",
    "keyId":4294968019,
    "condition":[
      "products.Color = ? AND products.Qty = ?"
    ],
    "columnstore_filter_predicates":[
      {
        "executor":"ColumnStoreFilterPredicate",
        "keyId":4294968020,
        "condition":[
          "products.Color = ? AND products.Qty = ?"
        ],
        "columnstore_filter_callback_ordinal":"2",
        "columnstore_filter_execution_type":"regular",
        "columnstore_filter_total_rows_in":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "columnstore_filter_total_rows_out":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "columnstore_filter_encoded_rows_in":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "columnstore_filter_encoded_rows_out":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "inputs":[]
      },
      {
        "executor":"ColumnStoreFilterPredicate",
        "keyId":4294968020,
        "condition":[
          "        products.Color = ?"
        ],
        "columnstore_filter_callback_ordinal":"3",
        "columnstore_filter_execution_type":"regular",
        "columnstore_filter_total_rows_in":{ "value":6, "avg":0.750000, "stddev":1.984313, "max":6, "maxPartition":2 },
        "columnstore_filter_total_rows_out":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
        "columnstore_filter_encoded_rows_in":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "columnstore_filter_encoded_rows_out":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
        "inputs":[]
      },
      {
        "executor":"ColumnStoreFilterPredicate",
        "keyId":4294968020,
        "condition":[
          "    AND products.Qty = ?"
        ],
        "columnstore_filter_callback_ordinal":"5",
        "columnstore_filter_execution_type":"regular",
        "columnstore_filter_total_rows_in":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
        "columnstore_filter_total_rows_out":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
        "columnstore_filter_encoded_rows_in":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
        "columnstore_filter_encoded_rows_out":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
        "inputs":[]
      }
    ],
    "subselects":[],
    "actual_row_count":{ "value":4, "avg":0.500000, "stddev":1.322876, "max":4, "maxPartition":2 },
    "actual_total_time":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
    "start_time":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":2 },
    "columnstore_filter_total_rows_in":{ "value":6, "avg":0.750000, "stddev":1.984313, "max":6, "maxPartition":2 },
    "columnstore_filter_avg_filters_per_row":{ "value":0, "avg":1.666667, "stddev":0.000000, "max":0, "maxPartition":0 },
    "columnstore_filter_avg_index_filters_per_row":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
    "columnstore_filter_avg_bloom_filters_per_row":{ "value":0, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
...

Example 6

Here is an example of the spilling related metrics under hash Join operators:

JSON

...
"inputs":[
    {
        "executor":"HashJoin",
        "keyId":458753,
        "type":"inner",
        "subselects":[],
        "actual_row_count":{ "value":4194304, "avg":1398101.333333, "stddev":606.633516, "max":1398899, "maxPartition":0 },
        "actual_total_time":{ "value":1964, "avg":1964.000000, "stddev":0.000000, "max":1964, "maxPartition":0 },
        "start_time":{ "value":7074, "avg":7486.333333, "stddev":0.000000, "max":8142, "maxPartition":2 },
        "inputs":[
            {
                "executor":"HashTableProbe",
                "keyId":458917,
                "condition":[
                    "r1.a = r0.a"
                    ],
                "spill_outputted_rows":{ "value":4194304, "avg":1398101.333333, "stddev":606.633516, "max":1398899, "maxPartition":0 },
                "spill_disk_usage":{ "value":134217728, "avg":44739242.666667, "stddev":19412.272498, "max":44764768, "maxPartition":0 },
                "inputs":[
                    {
                        "executor":"HashTableBuild",
                        "keyId":458916,
                        "alias":"r1",
                        "actual_row_count":{ "value":4194304, "avg":0.000000, "stddev":0.000000, "max":0, "maxPartition":0 },
                        "actual_total_time":{ "value":4, "avg":4.000000, "stddev":0.000000, "max":4, "maxPartition":0 },
                        "start_time":{ "value":1782, "avg":1866.666667, "stddev":0.000000, "max":1978, "maxPartition":1 },
                        "memory_usage":{ "value":393216, "avg":131072.000000, "stddev":0.000000, "max":131072, "maxPartition":0 },
                        "spill_outputted_rows":{ "value":4194304, "avg":1398101.333333, "stddev":606.633516, "max":1398899, "maxPartition":0 },
                        "spill_disk_usage":{ "value":100663296, "avg":33554432.000000, "stddev":14559.204374, "max":33573576, "maxPartition":0 },
...

On this page

Syntax

Remarks

Profiling a Hung Query

`PROFILE` Metrics

`PROFILE` Metrics for Columnstore Queries that Use Filters

`PROFILE` Blob Cache Metrics

`PROFILE JSON` Fields

`PROFILE JSON` Metrics with `profile_for_debug` = `ON`

`PROFILE JSON` Metrics with `enable_spilling` = `ON`

Visual Profile via SHOW PROFILE UI

Examples

Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

Was this article helpful?

On this page

Was this article helpful?

PROFILE

On this page

Syntax

Remarks

Profiling a Hung Query

PROFILE Metrics

PROFILE Metrics for Columnstore Queries that Use Filters

PROFILE Blob Cache Metrics

PROFILE JSON Fields

PROFILE JSON Metrics with profile_for_debug = ON

PROFILE JSON Metrics with enable_spilling = ON

Visual Profile via SHOW PROFILE UI

Examples

Example 1

Example 2

Example 3

Example 4

Example 5

Example 6

Related Topics

Was this article helpful?

On this page

Was this article helpful?

`PROFILE` Metrics

`PROFILE` Metrics for Columnstore Queries that Use Filters

`PROFILE` Blob Cache Metrics

`PROFILE JSON` Fields

`PROFILE JSON` Metrics with `profile_for_debug` = `ON`

`PROFILE JSON` Metrics with `enable_spilling` = `ON`