Thanks for the pointers Yan! Renato M.
2010/10/27 Yan Zhou <[email protected]> > If you can not change your input data generation process to generate > input directly in Zebra, I can’t see any alternative than two sets of data. > > > > Regarding generating Zebra data, Pig is simpler than raw map/reduce and the > performance should be fine too, provided there is a PIG loader for your > input data format. > > > > Yan > > > ------------------------------ > > *From:* Renato Marroquín Mogrovejo [mailto:[email protected]] > *Sent:* Wednesday, October 27, 2010 9:29 AM > *To:* Yan Zhou; [email protected] > *Subject:* Re: Using data with Zebra > > > > Thanks Yan! > > Just a couple of questions. The thing is that I have too much data just to > delete it and reprocess it all, and if I would reprocess all my hdfs data, > then I will generate the same amount of data duplicated, one with Zebra and > one with regular hdfs data. What would be the best approach that you would > suggest? and would it be better to use Pig or raw MapReduce? > > Renato M. > > 2010/10/25 Yan Zhou <[email protected]> > > .schema is column group's schema file; .btschema is Zebra table's schema > file; .meta is column group's index file. > > The bottom line is that they are all internal files maintained by Zebra and > users should not access or manipulate them directly. Also, the storage > format by Zebra is probably different from that used by you data already on > HDFS. > > In summary, you have to use Zebra to generate Zebra data and no other data > format can be used by Zebra. > > Yan > > > -----Original Message----- > From: Renato Marroquín Mogrovejo [mailto:[email protected]] > Sent: Sunday, October 24, 2010 1:15 PM > To: [email protected] > Subject: Using data with Zebra > > Hi there, I have some doubts about zebra usage. > The thing is that all my data is already in HDFS, and want to use the zebra > storers and loaders, but I don't want to reprocess all my data just to get > the .meta, .schema and the .btschema files, and by the way how are those > files related? I mean they all keep file's metadata, right? > Is there any way I can create the necessary files to use zebra's loaders > and > storers functionality? Any advice or suggestion is highly appreciated. > Thanks in advanced. > > > Renato M. > > >
