[ https://issues.apache.org/jira/browse/YARN-3411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14545788#comment-14545788 ]
Junping Du commented on YARN-3411: ---------------------------------- Hi [~vrushalic] and all, there is one question I had for a long time: Looks like we are updating time series metrics data as versions of metrics cell instead of inserting of a row (with time serve as a column). Do we have an estimation for performance cost for aggregation metrics data over cell versions vs. rows? The original usecase for cell versions coming from BigTable is to store web pages which could only be limited number of versions (too stale versions could be dropped). In our case, the metrics data could be updated much more frequently and we cannot drop the earliest data. I see we are setting max version to be 200, in "cf3.setMaxVersions(EntityTableDetails.ENTITY_TABLE_METRICS_MAX_VERSIONS_DEFAULT);". It sounds like it may not be enough for cases, like: long running services, container reuse, etc., this even sounds challenge for normal container (assume appCollector doesn't cache metrics data locally, the interval default is: 1 sec). Thoughts? > [Storage implementation] explore the native HBase write schema for storage > -------------------------------------------------------------------------- > > Key: YARN-3411 > URL: https://issues.apache.org/jira/browse/YARN-3411 > Project: Hadoop YARN > Issue Type: Sub-task > Components: timelineserver > Reporter: Sangjin Lee > Assignee: Vrushali C > Priority: Critical > Attachments: ATSv2BackendHBaseSchemaproposal.pdf, > YARN-3411-YARN-2928.001.patch, YARN-3411-YARN-2928.002.patch, > YARN-3411.poc.2.txt, YARN-3411.poc.3.txt, YARN-3411.poc.4.txt, > YARN-3411.poc.5.txt, YARN-3411.poc.6.txt, YARN-3411.poc.7.txt, > YARN-3411.poc.txt > > > There is work that's in progress to implement the storage based on a Phoenix > schema (YARN-3134). > In parallel, we would like to explore an implementation based on a native > HBase schema for the write path. Such a schema does not exclude using > Phoenix, especially for reads and offline queries. > Once we have basic implementations of both options, we could evaluate them in > terms of performance, scalability, usability, etc. and make a call. -- This message was sent by Atlassian JIRA (v6.3.4#6332)