In the middle of April 2020, UCloud cloud database product line released the MySQL version of Kuaijie UDB. As the latest generation of cloud database after the UDB product architecture upgrade, Kuaijie UDB adopts the mainstream computing storage separation architecture in the industry: The computing layer uses high-performance UCloud Kuaijie cloud host, and the storage layer uses ultra-high performance RSSD cloud disk, which is suitable for the vast majority of user scenarios, including the Internet, the Internet of Things, retail e-commerce, logistics, games and other industries.

As for the performance of Kuaijie cloud host, it has been tested and compared in detail in the “Ali cloud, Tencent cloud, UCloud, Huawei cloud and cloud host comparative test report”, and its support ability to MySQL database is particularly outstanding. On the basis of the underlying hardware of Kuaijie cloud host and RDMA NVME SSD cloud disk, the UCloud database team also made a series of customization and tuning work specifically for MySQL database operating system, so that Kuaijie version of MySQL UDB has unique performance advantages in different scenarios.

In order to facilitate the intuitive feeling of Kuaijie UDB performance in the real scene, the UDB team recently carried out a full range of performance comparison and testing. In this test, a highly available mysql5.7db environment was deployed, and the test was performed for 30 minutes using the Sysbench test tool. In addition, we selected three typical data volume scenarios for simulation test respectively. Ordinary version of UDB, Youshang RDS, Kuaijie UDB and self-built DB of Kuaijie cloud host were selected for each scenario as comparison objects.

Scenario 1: Small amount of data (in-memory)



This scenario mainly simulates the business scenario where all the data can be read into the cache for query. The figure above shows the QPS/TPS of 250 tables with 25000 rows of records per table (all the data in the memory cache during the test).

Scenario 2: Medium Data Volume (Medium IO type)





This scenario mainly simulates the DB usage scenario in which the business needs to read and write disk update cache to a certain extent. The figure above shows QPS/TPS with a data volume of 150 tables and 800,000 (800,000) rows per table (part of the data during the test is in the memory cache, and part of the data needs to be read and written from disk).

Scenario 3: Large data volume (large IO type)





This scenario mainly simulates the DB usage scenario where the business requires a large number of read and write disk update cache. The figure above shows a data volume of 150 tables with 8000000 (8 million) rows per table (most of the data during the test need to be read and written from disk) QPS/TPS.

Results analysis

From the test results, we can see that Kuaijie UDB has excellent performance in three different data volume scenarios, especially in scenarios with small data volume and medium data volume.



(Figure shows the average QPS for the three scenarios)

• Contrast fast Jieyun host self-built

The QPS performance of Kuaijie UDB is 30% higher on average. The main reason is that Kuaijie UDB has carried out further tuning of DB system and improvement of operating system on the basis of Kuaijie cloud host, adopted the latest and most stable firmware, and reduced the fluctuation influence of cloud hard disk IOPS through fine scheduling. These deep integration and tuning of the UCloud cloud platform ensure the performance of Kuaijie UDB.

• Compared with the old architecture, the performance of Kuaidjie UDB is better than that of the old architecture in most scenarios and configurations, and its QPS has increased by 24% on average, which is due to the stronger hardware configuration in the underlying Kuaidjie UDB after the architecture upgrade.

• The QPS performance of Kuaijie UDB is 32% higher on average than that of You-business database with the same configuration, which is due to the obvious performance advantage of UCloud Kuaijie cloud host platform itself in the industry (refer to the Comparison Test Report of Ali Cloud, Tencent Cloud, UCloud and Huawei Cloud Host). On the other hand, it benefits from the customization and optimization of Kuaijie UDB at the product level as mentioned above, which makes its performance advantages more prominent.

Quick jie UDB besides have obvious advantages in performance, also is the price for the user to maximize benefits, as shown in the figure below, we combine quick jie UDB, quick JieYun host self-built DB, bad performance data, and the product price, the price of quantitative (= the average ratio of QPS/offer), results show that the quick jie UDB highest performance to price ratio. It is 28.6% higher than self-built DB on average, 75% higher than friendly business RDS on average, and 9% higher than ordinary version SSD-UDB on average, so it is especially suitable for Internet e-commerce, games and other business scenarios that need more cost-effective DB services.

Kuaijie UDB advantage

After MySQL is upgraded to the new version of Kuaijie UDB, it is displayed on the console in the form of “NVME model” :



In addition to supporting the most mainstream MySQL 5.6 and MySQL 5.7, and seamlessly compatible with all the core functions of current UDB instances of the same version, users can also experience the high availability, high performance, high scalability, simplicity, low cost and other features brought by cloud database on Kuaijie UDB.

High availability: the bottom layer of KuaijieUDB adopts dual main hot standby architecture. The data of two nodes of the high availability instance will be synchronized in real time. When the main node fails to access, it will automatically switch to the standby node, completely solving the database unavailability caused by downtime or hardware failure. High availability dual master nodes can be deployed in the same available zone or in different available zones to support disaster recovery at the available zone level.

High performance: The storage layer RSSD cloud disk of Kuaijie UDB is based on the new generation of distributed block storage architecture. The bottom layer uses NVME SSD as the storage medium, and the network transmission uses RDMA, which can provide users with the random read and write capacity of up to 1.2 million on a single disk and lower single-channel delay capacity.

High scalability: users can flexibly upgrade and downgrade disks and memory online for Kuaijie UDB, and support large-capacity storage with the maximum capacity up to 32T, which is very suitable for large-capacity and high-performance business scenarios.

Simple and easy to use: MySQL version of Kuaijie UDB can help users save the self-built database work such as procurement, deployment and configuration through fast online deployment. It also provides functions such as automatic backup, data second-level file back, log management, automatic monitoring alarm, performance optimization, Clouddba and so on, making database operation and maintenance management more simple and convenient.

Low cost: users can immediately deploy database resources according to business needs, without the need to purchase high-cost servers in the early stage of the business, which effectively reduces one-time large-scale cost investment in the early stage. Flexible payment model can also avoid idle waste of resources.

conclusion

With the increasing amount of data in enterprises, the improvement of database performance becomes particularly important. After the architecture and hardware upgrade, Kuaijie UDB’s high performance and high cost performance are the excellent choice for enterprises to deploy high-performance databases, whether compared with the self-built DB or the RDS with the same configuration from You-Business.

At present, KuaijieUDB has been launched in Beijing Zone B, Beijing Zone C, Beijing Zone E, Shanghai Zone A, Guangzhou Zone B, Hong Kong, Taipei, Ho Chi Minh, Lagos, Seoul, and will be launched in the Philippines, Bangkok and other areas, to help users realize global deployment.

In addition, Kuaidi UDB will support MySQL5.5 and Percona 5.5/5.6/5.7 versions in September, to meet the needs of multiple versions of users. In terms of database types, it will also cover MongoDB, SQL Server and other more types, and is committed to creating a cloud database platform with high performance and availability for users.