[ https://issues.apache.org/jira/browse/MESOS-3765?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14993646#comment-14993646 ]
Adam B commented on MESOS-3765: ------------------------------- We would definitely consider role weights. A role's fair share is calculated based on the number of active roles (with registered frameworks) and their weights. With 4 roles each with a weight of 1, each role's fair share of the cluster is 25% of the total cpu/mem/disk resources. If one of those roles has 5 frameworks in it, then each of those frameworks' fair share is 5% of the cluster resources. (There are no per-framework weights.) Quota would be global rather than per-node, but even with a global quota, it may be subdivided so that a framework can say "I need 5 instances of {2cpu, 4GB RAM}, although I don't care which nodes those instances come from." As such, we would use {2cpu, 4GB RAM} as the minimum per-node offer to a framework, rather than offering the entire node. See MESOS-1791 for more information about quota. > Make offer size adjustable (granularity) > ---------------------------------------- > > Key: MESOS-3765 > URL: https://issues.apache.org/jira/browse/MESOS-3765 > Project: Mesos > Issue Type: Improvement > Components: allocation > Reporter: Alexander Rukletsov > Assignee: Guangya Liu > > The built-in allocator performs "coarse-grained" allocation, meaning that it > always allocates the entire remaining agent resources to a single framework. > This may heavily impact allocation fairness in some cases, for example in > presence of numerous greedy frameworks and a small number of powerful agents. > A possible solution would be to allow operators explicitly specify > granularity via allocator flags. While this can be tricky for non-standard > resources, it's pretty straightforward for {{cpus}} and {{mem}}. -- This message was sent by Atlassian JIRA (v6.3.4#6332)