Thanks for the clarification. I'm happily on board with this as long as our approach takes account of the differences between execution engines. While MR and Tez might be similar, there could be new execution engines in the future which might not be that similar. Ideally, all execution engines should benefit from this effort yet room is kept to allow for specific optimizations for a particular engine. It's great if we all see that.
Thanks, Xuefu On Mon, Jun 23, 2014 at 11:05 AM, John Pullokkaran < jpullokka...@hortonworks.com> wrote: > Following may help in reducing the confusion: > > 1. In design doc the cost formula is for choosing Join Algorithm. The cost > formula as described in the doc assumes Tez execution. > > 2. However current work on CBO doesn’t include Join algorithm selection. > Instead it rearranges Join based on Join cardinality & NDV. In other words > Join reordering is not depended on Physical Execution Layer (Tez or MR). > > 3. When we decide to do Join Algorithm Selection we can fit in cost formula > for both a) MR b) Tez. This way, based on the physical execution layer we > can select best Join Algorithm/Order. > > 4. The cost formula for Join Algorithm selection is not that different > between MR & Tez (except for intermediate HDFS writes). So assume that CBO > can support both execution layers rather easily. > > 5. CBO framework allows you to plug and play any cost model. There is no > hard coupling. > > Thanks > John > > > > On Mon, Jun 23, 2014 at 7:09 AM, Xuefu Zhang <xzh...@cloudera.com> wrote: > > > Hi Ashutosh, > > > > Thanks for your information. I do have some questions, but my concern is > > more on the design doc than branching. Nevertheless, I think it would be > > very helpful to clarify in the design before we actually put a lot of > > effort. > > > > From the design doc, it seems that the cost estimation is based on Tez, > > while the optimization occurs on logical layer. I'd think that CBO are > > valuable to either engine. If there is anything that's specific to a > > particular to an engine, then that optimization should stay at engine > > layer. > > > > My original comments was posted on HIVE-5775. Please let me know your > > thoughts. I'd also like to hear from the community. > > > > > > > https://issues.apache.org/jira/browse/HIVE-5775?focusedCommentId=14039987&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14039987 > > > > Thanks, > > Xuefu > > > > > > On Thu, Jun 19, 2014 at 10:34 PM, Ashutosh Chauhan <hashut...@apache.org > > > > wrote: > > > > > Hi all, > > > > > > Some of you may have noticed that cost based optimizer work is going on > > at > > > HIVE-5775 John has put up an initial patch there as well. But there is > a > > > lot more work that needs to be done. Following our tradition of large > > > feature work in branch, I propose that we create a branch and commit > this > > > patch in it and than continue to work on it in branch to improve it. > > > Hopefully, we can get it in shape so that we can merge it in trunk once > > its > > > ready. Unless, I hear otherwise I plan to create a branch and commit > > this > > > initial patch by early next week. > > > > > > > > > Design doc is located here : > > > > > > > > > https://cwiki.apache.org/confluence/display/Hive/Cost-based+optimization+in+Hive > > > > > > > > > Thanks, > > > > > > Ashutosh > > > > > > > -- > CONFIDENTIALITY NOTICE > NOTICE: This message is intended for the use of the individual or entity to > which it is addressed and may contain information that is confidential, > privileged and exempt from disclosure under applicable law. If the reader > of this message is not the intended recipient, you are hereby notified that > any printing, copying, dissemination, distribution, disclosure or > forwarding of this communication is strictly prohibited. If you have > received this communication in error, please contact the sender immediately > and delete it from your system. Thank You. >
