Return-Path: X-Original-To: apmail-hive-dev-archive@www.apache.org Delivered-To: apmail-hive-dev-archive@www.apache.org Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by minotaur.apache.org (Postfix) with SMTP id 6572517BB8 for ; Sat, 11 Oct 2014 00:39:35 +0000 (UTC) Received: (qmail 25532 invoked by uid 500); 11 Oct 2014 00:39:34 -0000 Delivered-To: apmail-hive-dev-archive@hive.apache.org Received: (qmail 25398 invoked by uid 500); 11 Oct 2014 00:39:33 -0000 Mailing-List: contact dev-help@hive.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@hive.apache.org Delivered-To: mailing list dev@hive.apache.org Received: (qmail 25062 invoked by uid 500); 11 Oct 2014 00:39:33 -0000 Delivered-To: apmail-hadoop-hive-dev@hadoop.apache.org Received: (qmail 25021 invoked by uid 99); 11 Oct 2014 00:39:33 -0000 Received: from arcas.apache.org (HELO arcas.apache.org) (140.211.11.28) by apache.org (qpsmtpd/0.29) with ESMTP; Sat, 11 Oct 2014 00:39:33 +0000 Date: Sat, 11 Oct 2014 00:39:33 +0000 (UTC) From: "Gopal V (JIRA)" To: hive-dev@hadoop.apache.org Message-ID: In-Reply-To: References: Subject: [jira] [Updated] (HIVE-8349) DISTRIBUTE BY should work with tez auto-parallelism enabled MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-JIRA-FingerPrint: 30527f35849b9dde25b450d4833f0394 [ https://issues.apache.org/jira/browse/HIVE-8349?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Gopal V updated HIVE-8349: -------------------------- Component/s: Tez > DISTRIBUTE BY should work with tez auto-parallelism enabled > ----------------------------------------------------------- > > Key: HIVE-8349 > URL: https://issues.apache.org/jira/browse/HIVE-8349 > Project: Hive > Issue Type: Bug > Components: Physical Optimizer, Tez > Affects Versions: 0.14.0 > Reporter: Gopal V > Assignee: Gopal V > Attachments: HIVE-8349.1.patch > > > Current implementation of DISTRIBUTE BY does not work when tez auto-parallelism is turned on, because of hashCode distribution issues. > In case of distribute by, the key is actually zero bytes, with only partitioning enabled via hashCode - this adversely affects the uniform hashing implementation. > In an ideal scenario, the edge should go from the ordered kv input to the unordered partitioned edge, to speed up the processing massively. -- This message was sent by Atlassian JIRA (v6.3.4#6332)