Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 3.10 reconnections (per second) General statistics: total time: 120.0463s # total time total number of events: 1397185 # toal number of transactions Latency (ms): min: 5.37 # minimum latency avg: 17.13 time of over 95th percentile. sum: 23999546.19 Threads fairness: events (avg/stddev): 6985.9250/34.74 # On average, 6985.9250 events were completed per thread, and the standard deviation is 34.74 execution

Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 8.10 CDC can monitor data changes in the storage nodes of ShardingSphere‐Proxy, capture data operation events, filter and extract useful infor‐ mation, and finally send these changed data to a specified target 300ms). When processing XA transactions, if the received multiple physical database incremental events exceed 300ms, it may cause the XA transaction to be split and pushed. Handling of large transactions

Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 3.10 reconnections (per second) General statistics: total time: 120.0463s # total time total number of events: 1397185 # toal number of transactions Latency (ms): min: 5.37 # minimum latency avg: 17.13 time of over 95th percentile. sum: 23999546.19 Threads fairness: events (avg/stddev): 6985.9250/34.74 # On average, 6985.9250 events were completed per thread, and the standard deviation is 34.74 execution

Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 8.10 reconnections (per second) General statistics: total time: 120.0463s # total time total number of events: 1397185 # toal number of transactions 11.6. Performance Test 470 Apache ShardingSphere document time of over 95th percentile. sum: 23999546.19 Threads fairness: events (avg/stddev): 6985.9250/34.74 # On average, 6985.9250 events were completed per thread, and the standard deviation is 34.74 execution

important events in the scheduling process through event subscription for query, statistics and monitor. Now, the event subscription based on relation database is provided to record events, and developers

underly‐ ing concurrency engine and send events according to each key steps during the execution process, such as starting, successful and failed execution events. Execution engine only focuses on message ShardingSphere document, v5.0.0-beta tions, invoked chain tracing and so on, will subscribe focusing events and do corresponding operations. Through the connection mode acquired in preparation phrase, ShardingSphere logic schemas, it always hold the two schemas and will be notified about the table/rule changed events in the two schemas. 2. Since version 4.1.0, sharing‐proxy support adding a new logic schema dynamically

Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 4.10 underly‐ ing concurrency engine and send events according to each key steps during the execution process, such as starting, successful and failed execution events. Execution engine only focuses on message ShardingSphere document, v5.1.1 tions, invoked chain tracing and so on, will subscribe focusing events and do corresponding operations. Through the connection mode acquired in preparation phrase, ShardingSphere

Challenges Tracing and metrics need to collect system information through event tracking. Lots of events tracking make kernel code mess, difficult to maintain, and difficult to customize extend. 4.10 underly‐ ing concurrency engine and send events according to each key steps during the execution process, such as starting, successful and failed execution events. Execution engine only focuses on message ShardingSphere document, v5.1.2 tions, invoked chain tracing and so on, will subscribe focusing events and do corresponding operations. Through the connection mode acquired in preparation phrase, ShardingSphere

underly‐ ing concurrency engine and send events according to each key steps during the execution process, such as starting, successful and failed execution events. Execution engine only focuses on message ShardingSphere document, v5.0.0 tions, invoked chain tracing and so on, will subscribe focusing events and do corresponding operations. Through the connection mode acquired in preparation phrase, ShardingSphere

number of events: 1397185 # 总共发生多 少事务数 Latency (ms): min: 5.37 # 最小延时 avg: 17.13 # 平均延时 max: 109.75 # 最大延时 95th percentile: 24.83 # 超过 95% 平均耗时 sum: 23999546.19 Threads fairness: events (avg/stddev):