I seem to recall J-D mentioning recently that they use a single character for their column family names. This is implicit in Section 6.4 (Try to minimize row and column sizes), but you could maybe expand on the point a little with some recommended actions. Reducing the size of the column family name is much easier than reducing column qualifier name (as those are not always under the designers control). It is a little counter-intuitive as a lot of popular press emphasizes that "a couple of bytes do not matter". However, when you store it in very single cell, it starts to add up -- even with compression.
Just something I wish I would have understood back when I last designed an hbase schema. Dave -----Original Message----- From: Doug Meil [mailto:[email protected]] Sent: Saturday, August 13, 2011 6:21 PM To: [email protected] Subject: Re: Column family limitations See the Hbase book... http://hbase.apache.org/book.html#number.of.cfs On 8/13/11 8:17 PM, "Mark" <[email protected]> wrote: >I don't quite remember where but I think I remember hearing that you >should only have a handful of column families (1-4) per table. Is this >true and if so, why? > >Thanks
