[ https://issues.apache.org/jira/browse/SPARK-1767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13993909#comment-13993909 ]
Sandy Ryza commented on SPARK-1767: ----------------------------------- Currently, RDDs only support a single level of location preference through RDD#preferredLocations(split), which returns a sequence of strings. To prefer cached-replicas, this needs to be extended in some way. We could deprecate preferredLocations and add a preferredLocations(split, storageType), where storageType is MEMORY, DISK, and eventually FLASH? Maybe more hackily, we could give the location strings a prefix like "inmem:" that specifies the storage type. > Prefer HDFS-cached replicas when scheduling data-local tasks > ------------------------------------------------------------ > > Key: SPARK-1767 > URL: https://issues.apache.org/jira/browse/SPARK-1767 > Project: Spark > Issue Type: Improvement > Components: Spark Core > Affects Versions: 1.0.0 > Reporter: Sandy Ryza > -- This message was sent by Atlassian JIRA (v6.2#6252)