[ https://issues.apache.org/jira/browse/KYLIN-2731?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhong Yanghong updated KYLIN-2731: ---------------------------------- Description: The basic idea of cube optimization without influencing current queries is shown at figure [^One-Click Cube Optimization.png]. To optimize a cube with N ready segments, * For each segment, a corresponding optimize job will be triggered to adjust its cuboids. * If a optimize job finished, it's resulting segment status should be READY_PENDING and will not serve query. * There should be job to wait for all of the optimize jobs finishing their work, and then atomically remove old READY segments and update the status of READY_PENDING segments to READY for serving query. Here, we call this kind of job as {color:#f79232}checkpoint executable{color}. was: The basic idea of cube optimization without influencing current queries is shown at figure [^One-Click Cube Optimization.png]. To optimize a cube with N ready segments, * For each segment, a corresponding optimize job will be triggered to adjust its cuboids. * If a optimize job finished, it's resulting segment status should be READY_PENDING and will not serve query. * There should be job to wait for all of the optimize jobs finishing their work, and then atomically remove old READY segments and update the status of READY_PENDING segments to READY for serving query. > Introduce checkpoint executable > ------------------------------- > > Key: KYLIN-2731 > URL: https://issues.apache.org/jira/browse/KYLIN-2731 > Project: Kylin > Issue Type: Sub-task > Affects Versions: v2.2.0 > Reporter: Zhong Yanghong > Assignee: Zhong Yanghong > Attachments: One-Click Cube Optimization.png > > > The basic idea of cube optimization without influencing current queries is > shown at figure [^One-Click Cube Optimization.png]. > To optimize a cube with N ready segments, > * For each segment, a corresponding optimize job will be triggered to adjust > its cuboids. > * If a optimize job finished, it's resulting segment status should be > READY_PENDING and will not serve query. > * There should be job to wait for all of the optimize jobs finishing their > work, and then atomically remove old READY segments and update the status of > READY_PENDING segments to READY for serving query. Here, we call this kind of > job as {color:#f79232}checkpoint executable{color}. -- This message was sent by Atlassian JIRA (v6.4.14#64029)