The HFile "index" is an implementation detail, it doesnt affect how the top level API presents itself. At the highest level HBase provides an API that lets you store sorted order keys, and then seek to a random one and start scanning, or just get one row. With those semantics, building a business logic index on that is trivial.
On Wed, Nov 11, 2009 at 11:45 PM, jdh <tianqiu_...@yahoo.com.cn> wrote: > > Hi all ! > The Bigtable paper related little to index. I only know hfile is similar to > mapfile in hadoop, and there are two files in mapfile: data and index. But > still I have no idea about how the index is implemented in hbase. Can > somebody give me any advice? > Thanks! > Regards! > yan > -- > View this message in context: > http://old.nabble.com/about-index-in-HBase-tp26314679p26314679.html > Sent from the HBase User mailing list archive at Nabble.com. > >
