Hello,
> Could you tell me about internals or why jdbc shows more good performance? The deployment architecture of jdbc and proxy can be found in [1]. In general, jdbc is a driver embedded in your APP (similar to the mysql driver), so the request is directly from the app to the database cluster. While proxy is deployed as a Sidecar on a pod or on a server, each request from the APP is forwarded through proxy to your database cluster (Storage cluster of ShardingSphere distributed DB system). This case will result in additional TPC overhead (Same as other distributed DB system). > Just found this: https://github.com/apache/shardingsphere/pull/21283 SQL hint is another useful feature for some user arbitrary visiting cases. [1] https://shardingsphere.apache.org/document/current/en/overview/ ------------------------------------------------------------------------------ Trista Pan Twitter & Github: tristaZero On 10/5/2022 08:10，Edward J. Yoon<edwardy...@apache.org> wrote： Just found this: https://github.com/apache/shardingsphere/pull/21283 On Tue, Oct 4, 2022 at 8:35 PM Edward J. Yoon <edwardy...@apache.org> wrote: Thanks Juan. I saw some blog posts that said shardingsphere is about a 20~30% performance penalty compared with direct connections. Could you tell me about internals or why jdbc shows more good performance? On Tue, Oct 4, 2022 at 8:16 PM Juan Pan <panj...@apache.org> wrote: Jdbc is 30-40% higher than proxy based on different scenarios [1]. However, since `proxy + mysql` is similar with other distributed database architecture, that means proxy has the similar performance with other sharding databases and is ready for production env (from the feedback of our community [2, 3]). [1] https://shardingsphere.apache.org/document/current/en/test-manual/performance-test/benchmarksql-proxy-sharding-test/ [2] https://shardingsphere.apache.org/blog/en/material/ [3] https://www.mysql.com/customers/view/?id=1461 ------------------------------------------------------------------------------ Trista Pan Twitter & Github: tristaZero On 10/4/2022 19:05，Edward J. Yoon<edwardy...@apache.org> wrote： Just curious about whether there's some performance difference between jdbc vs. proxy. On Tue, Oct 4, 2022 at 8:01 PM Edward J. Yoon <edwardy...@apache.org> wrote: Currently there are 4 set consists of single master and single slave, and one integrated db based on multi source replication. % 4 modular sharding is used at application level. So, I'm looking for the elastic scalability, easy-to-use/maintenance, and the high performance middleware like shardingsphere. On Tue, Oct 4, 2022 at 7:50 PM Juan Pan <panj...@apache.org> wrote: Hello Edward J. Yoon, adopting shardingshpere for the order database that requires to cover heavy read and write. Does that mean you want ShardingSphere to do the read-write-splitting function? sharded mysql cluster with division method You already sharded your mysql database cluster? Do you plan to let ShardingSphere help you manage all the mysql shards? ShardingSphere currently offers two clients: ShardingSphere-JDBC (for Java APP) and ShardingSphere-Proxy. In your case, I recommend deploying ShardingSphere-Proxy[1] to help you with read-write-splitting or sharding management. Then, for your Python workload, simply connect to ShardingSphere-Proxy using the MySQLdb library. Best wishes, Trista [1] https://shardingsphere.apache.org/document/current/en/quick-start/shardingsphere-proxy-quick-start/ ------------------------------------------------------------------------------ Trista Pan Twitter & Github: tristaZero On 10/4/2022 18:29，Edward J. Yoon<edwardy...@apache.org> wrote： Hello, I am edwardyoon, currently working at Delivery Hero. From there, I am considering adopting shardingshpere for the order database that requires to cover heavy read and write. Our system is consists of django and sharded mysql cluster with division method at this moment. I’d like to know whether there is a reliable django library that can be used for real production. Thanks. -- Best Regards, Edward J. Yoon -- Best Regards, Edward J. Yoon -- Best Regards, Edward J. Yoon -- Best Regards, Edward J. Yoon -- Best Regards, Edward J. Yoon
