SelectDB Topped ClickBench - a Benchmark For Analytical DBMS

It runs on c6a.4xlarge VM in AWS with 500 GB gp2 by default, surpassing a number of global top products by leading in a lot of indicators.

ABOUT CLICKBENCH

In ClickBench, the test data are all taken from the actual traffic recording of one of the world's largest web analytics platforms, covering a variety of data types and typical scenarios such as ad hoc queries and statistical reports, which can truly reflect the performance of major databases in the production environment, thus attracting many internationally renowned databases like Snowflake, Redshift, Athena, Greenplum, Druid ,etc..

The measured indicators include the time taken to import the same dataset, the amount of storage space occupied, and the time taken to execute SQL under a specific machine, which are used to measure data import performance, data compression ratio, and query performance respectively. The one with the best performance in all test results will become the baseline. The indicators of the same test items will be compared with the baseline data and a ratio will be obtained. This ratio will reflect the gap with the best performance. When a new test result exceeds the original baseline, it will automatically become a new baseline.

As far as query performance, hot run and cold run will be executed for each SQL respectively to count the duration, that is, repeat the SQL for three times and select the one with the shortest duration, and execute directly after starting and cleaning the memory. Finally, the ratio of the execution time of all SQL to the baseline will be geometric averaged, which is the final test result. Therefore, ClickBench pays more attention to the excellent performance of the database in all test scenarios, rather than one or several scenarios, which makes the database require all-round capability improvement.

ABOUT SELECTDB

SelectDB is developed based on the commercial requirements of the open-source product Apache Doris. Without the support of Apache Doris community, our R&D speed would not be so fast. Therefore, we would like to thank all the technical partners of the Apache Doris community. All the performance optimization points for SelectDB will be returned to the Apache Doris community without any reservation. At present, some codes have been merged into the community master branch, and many PRs are in the process of merging. We believe that in the near future, all community users will be able to have the ultimate experience brought by the performance leap.

As a new generation of cloud-native data warehouse built from Apache Doris and running on multi-cloud, SelectDB has the core features of extremely fast, single unified, easy-to-use, and cost effective. It provides multi-cloud consistent experience and supports for most data types. Up to now, SelectDB has carried out in-depth cooperation with Alibaba Cloud, Tencent Cloud, AWS, and other well-known cloud service providers, and has serviced many enterprises in different industries.

THE RECORD OF SELECTDB

In the test results, SelectDB reached the top 1 with its strong query performance. Without any tuning, the query performance of SelectDB was No.1 on c6a.4xlarge VM in AWS with 500 GB gp2, and the performance of Hot Run and Cold Run were 35% and 25% ahead of the second respectively. In all 43 SQL statements, SelectDB got the best performance in nearly half of the query statements and become the new baselines. At the same time, the data loading efficiency of SelectDB ranked second among all products in the same machine. Before compression, 70G data writing only took 482s, and the single node writing speed exceeds 140MB/s. This ensures writing efficiency while achieving extreme query performance.

As far as query time consumption, compared with other top products, the overall query time consumption of SelectDB had decreased by 63%, 75%, 78%, 99%, and 459% respectively, and its performance is significantly ahead of similar products.

In the general list of a number of different machines and systems (including the world- famous cloud data warehouses such as Snowflake, Redshift, etc.), SelectDB still had the best Cold Run query performance and the second Hot Run query performance among all products of the same type, demonstrating a strong performance advantage.

In the test results running on the machine of c6a.metal, 500gb gp2 which was another common type, the overall query time of SelectDB in the Cold Run scenario was only 1/4 to 1/10 of that of similar products. Without any tuning, the highest query performance was nearly 11 times higher than that of similar products, showing a hugely leading edge.

SUMMARY

Today's achievement is one of the goals for SelectDB. In November, SelectDB Cloud will launch the 2.0 version. At that time, more wonderful technologies will be shared with users. Now, users are welcome to apply for a free trial on website (http://en.selectdb.com), the experts will provide professional service for users.

Hashtag: #SelectDB