ClickHouse/dbms/tests/performance
2019-09-20 23:26:01 +03:00
..
agg_functions_min_max_any.xml
array_element.xml
array_join.xml
base64_hits.xml
base64.xml
basename.xml
bounding_ratio.xml
cidr.xml
codec_double_delta.xml
codec_gorilla.xml
column_column_comparison.xml
columns_hashing.xml
complex_array_creation.xml
concat_hits.xml
conditional.xml
consistent_hashes.xml
constant_column_comparison.xml
constant_column_search.xml
count.xml
cpu_synthetic.xml
create_benchmark_page.py
cryptographic_hashes.xml
date_parsing.xml
date_time.xml
empty_string_deserialization.xml
empty_string_serialization.xml
entropy.xml
first_significant_subdomain.xml
fixed_string16.xml
float_parsing.xml
format_date_time.xml
functions_coding.xml
functions_geo.xml
general_purpose_hashes.xml
group_array_moving_sum.xml
h3.xml
if_array_num.xml
if_array_string.xml
if_string_const.xml
if_string_hits.xml
insert_values_with_expressions.xml
inserts_arrays_lowcardinality.xml
int_parsing.xml
IPv4.xml
IPv6.xml
jit_large_requests.xml
jit_small_requests.xml
joins_in_memory_pmj.xml
joins_in_memory.xml
json_extract_rapidjson.xml
json_extract_simdjson.xml
leftpad.xml
linear_regression.xml
math.xml
merge_table_streams.xml
merge_tree_many_partitions_2.xml
merge_tree_many_partitions.xml
merge_tree_simple_select.xml
ngram_distance.xml
number_formatting_formats.xml
nyc_taxi.xml
order_by_decimals.xml
order_by_read_in_order.xml
order_by_single_column.xml
parse_engine_file.xml
read_hits_with_aio.xml
README.md
right.xml
roundDown.xml
select_format.xml
set_hits.xml
set.xml
simple_join_query.xml
slices_hits.xml
sort.xml
string_sort.xml
system_numbers.xml
trim_numbers.xml
trim_urls.xml
trim_whitespace.xml
uniq.xml
url_hits.xml
visit_param_extract_raw.xml
website.xml

ClickHouse performance tests

This directory contains .xml-files with performance tests for clickhouse-performance-test tool.

How to write performance test

First of all you should check existing tests don't cover your case. If there are no such tests than you should write your own.

There two types of performance tests:

  • First is executed in loop, and have tag <type>loop</type> in config.
  • Second one is executed only once and have tag <type>once</type> in config.

Type once should be used only for endless queries. Even if your query really long (10 seconds+), it's better to choose loop test.

After you have choosen type, you have to specify preconditions. It contains table names. Only hits_100m_single, hits_10m_single, test.hits are available in CI.

The most important part of test is stop_conditions. For loop test you should always use min_time_not_changing_for_ms stop condition. For once test you can choose between average_speed_not_changing_for_ms and max_speed_not_changing_for_ms, but first is preferable. Also you should always specify total_time_ms metric. Endless tests will be ignored by CI.

metrics and main_metric settings are not important and can be ommited, because loop tests are always compared by min_time metric and once tests compared by max_rows_per_second.

You can use substitions, create, fill and drop queries to prepare test. You can find examples in this folder.

Take into account, that these tests will run in CI which consists of 56-cores and 512 RAM machines. Queries will be executed much faster than on local laptop.

How to run performance test

You have to run clickhouse-server and after you can start testing:

$ clickhouse-performance-test --input-file my_lovely_test1.xml --input-file my_lovely_test2.xml
$ clickhouse-performance-test --input-file /my_lovely_test_dir/