TPC-H Sample Data in the SQL Template

TPC-H Sample Data

TPC-H is a test set developed by the Transaction Processing Performance Council (TPC) to simulate decision-making support applications. It is widely used in academia and industry to evaluate the performance of decision-making support technology. This business test has higher requirements on vendors, because it can comprehensively evaluate the overall business computing capability. With universal business significance, is widely used in analysis of bank credit, credit card, telecom operation, tax, as well as tobacco industry decision-making analysis.

The TPC-H benchmark test is developed from TPC-D (a standard specified by TPC in 1994 and used as the test benchmark for decision-making support systems). TPC-H implements a 3NF data warehouse that contains eight basic relationships, with a data volume range from 1 GB to 3 TB. The TPC-H benchmark test includes 22 queries (Q1 to Q22). The main evaluation indicator is the response time of each query (from submission to result return). The unit of the TPC-H benchmark test is the query number per hour (QphH@size). H indicates the average number of complex queries per hour. size indicates the size of database, which reflects the query processing capability of the system. TPC-H can evaluate key performance parameters that other tests cannot evaluate, because it is modeled based on the actual production and operation environment. In a word, the TPC-H standard by TPC meets the test requirements of data warehouse and motivate vendors and research institutes to stretch the limit of this technology.

In this example, DLI directly queries the TPC-H dataset on OBS. DLI has generated a standard TPC-H-2.18 dataset of 100 MB which is uploaded to the tpch folder on OBS. The read-only permission is granted to you to facilitate query operations.

TPC-H Test and Metrics

TPC-H test is divided into three sub-tests: data loading test, Power test, and Throughput test. Data loading indicates the process of setting up a test database, and the loading test is to test the data loading ability of DBMS. The first test is data loading test that tests data loading time, which is time-consuming. The second test is Power test, also called raw query. After data loading test is complete, the database is in the initial state without any other operation, especially the data in the buffer is not tested. Power test requires that the 22 queries be executed once in sequence and a pair of RF1 and RF2 operations be executed at the same time. The third test is Throughput test, the core and most complex test, more similar to the actual application environment. With multiple query statement groups and a pair of RF1 and RF2 update flows, Throughput test pose greater pressure on the SUT system than Power test does.

Service Scenario

You can use the built-in TPC-H test suite of DLI to perform interactive query without uploading data.

Advantages of DLI Built-in TPC-H

• You can log in to DLI and get permission to run SQL statements without creating tables or import data.
• The 22 preset TPC-H SQL query templates with rich functions meet the requirements of most business scenarios. You do not need to download TPC-H query statements, which saves your time and energy.
• Data Lake gives you brand-new experience of serverless DLI product within the minimum time.

Precautions

When a sub-account uses the TPC-H test suite, the main account needs to grant the sub-account the OBS access permission and the permission to view the main account table. If the master account has not logged in to DLI, the sub-account needs to have the permissions to create databases and tables in addition to the preceding permissions.

Developing and Submitting a Spark SQL Job Using the TPC-H Sample Template

DLI allows you to create custom templates or save currently used SQL statements as templates for quick and convenient SQL operations. Once a template is saved, you can execute SQL operations directly through the template without the need to write SQL statements.

1. Log in to the DLI management console.
2. On the DLI management console, choose Job Templates > SQL Templates in the navigation pane on the left. On the displayed Sample Templates tab, click the plus sign next to tpchQuery to find the Q1_Price_summary_report_query template. Then, click Execute in the Operation column.

3. In the upper part of the editing window, set Engine to Spark, Queues to default, and Databases to default, and click Execute.

4. Check the query result on the View Result tab below the editing window.

This example uses the default queue and database preset in the system as an example. You can also run the command in a self-created queue and database.