Thanks Alan for this explanation. Interesting to see Primary Key in Hive.
Sometimes comparison is made between Hive Storage Index concept in Orc and Oracle Exadata storage index that also uses the same terminology! It is a bit of a misnomer to call Oracle Exadata indexes a “storage index”, since it appears that Exadata stores data block from tables in the storage index, usually when they are accessed via a full-table scan. In this context Exadata storage index is not a “real” index in the sense that the storage index exists only in RAM, and it must be re-created from scratch when the Exadata server is bounced. Oracle Exadata and SAP HANA as far as I know force serial scans into Hardware - with HANA, it is by pushing the bitmaps into the L2 cache on the chip - Oracle has special processors on SPARC T5 called D???? <something> that offloads the column bit scan off the CPU and onto separate specialized HW. As a result, both rely on massive parallelization.. Orc storage index is neat and different from both Exadata and SAP HANA, The way I see ORC storage indexes * They are combined Index and statistics. * Each index has statistics of min, max, count, and sum for each column in the row group of 10,000 rows. * Crucially, it has the location of the start of each row group, so that the query can jump straight to the beginning of the row group. * The query can do a SARG pushdown that limits which rows are required for the query and can avoid reading an entire file, or at least sections of the file which is by and large what a conventional RDBMS B-tree index does. Cheers, Dr Mich Talebzadeh LinkedIn https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw Sybase ASE 15 Gold Medal Award 2008 A Winning Strategy: Running the most Critical Financial Data on ASE 15 http://login.sybase.com/files/Product_Overviews/ASE-Winning-Strategy-091908.pdf Author of the books "A Practitioner’s Guide to Upgrading to Sybase ASE 15", ISBN 978-0-9563693-0-7. co-author "Sybase Transact SQL Guidelines Best Practices", ISBN 978-0-9759693-0-4 Publications due shortly: Complex Event Processing in Heterogeneous Environments, ISBN: 978-0-9563693-3-8 Oracle and Sybase, Concepts and Contrasts, ISBN: 978-0-9563693-1-4, volume one out shortly http://talebzadehmich.wordpress.com <http://talebzadehmich.wordpress.com/> NOTE: The information in this email is proprietary and confidential. This message is for the designated recipient only, if you are not the intended recipient, you should destroy it immediately. Any information in this message shall not be understood as given or endorsed by Peridale Technology Ltd, its subsidiaries or their employees, unless expressly so stated. It is the responsibility of the recipient to ensure that this email is virus free, therefore neither Peridale Technology Ltd, its subsidiaries nor their employees accept any responsibility. From: Alan Gates [mailto:alanfga...@gmail.com] Sent: 01 February 2016 17:07 To: user@hive.apache.org Subject: Re: ORC format ORC does not currently expose a primary key to the user, though we have talked of having it do that. As Mich says the indexing on ORC is oriented towards statistics that help the optimizer plan the query. This can be very important in split generation (determining which parts of the input will be read by which tasks) as well as on the fly input pruning (deciding not to read a section of the file because the stats show that no rows in that section will match a predicate). Either of these can help joins. But as there is not a user visible primary key there's no ability to rewrite the join as an index based join, which I think is what you were asking about in your original email. Alan. <mailto:philjj...@gmail.com> Philip Lee February 1, 2016 at 7:27 Also, when making ORC from CSV, for indexing every key on each coulmn is made, or a primary on a table is made ? If keys are made on each column in a table, accessing any column in some functions like filtering should be faster. -- ========================================================== Hae Joon Lee Now, in Germany, M.S. Candidate, Interested in Distributed System, Iterative Processing Dept. of Computer Science, Informatik in German, TUB Technical University of Berlin In Korea, M.S. Candidate, Computer Architecture Laboratory Dept. of Computer Science, KAIST Rm# 4414 CS Dept. KAIST 373-1 Guseong-dong, Yuseong-gu, Daejon, South Korea (305-701) Mobile) 49) 015-251-448-278 in Germany, no cellular in Korea ========================================================== <mailto:philjj...@gmail.com> Philip Lee February 1, 2016 at 7:21 Hello, I experiment the performance of some systems between ORC and CSV file. I read about ORC documentation on Hive website, but still curious of some things. I know ORC format is faster on filtering or reading because it has indexing. Has it advantage of joining two tables of ORC dataset as well? Could you explain about it in detail? When experimenting, it seems like it has some advantages of joining in some aspect, but not quite sure what characteristic of ORC make this happening rather than CSV. Best, Phil
