Hi,

Some of the experiments [1] show that using single column (single KV) allows 
for smaller file sizes and less data transfer to the client. On the other hand, 
natural split into columns (1 data attribute = 1 column) is somewhat "more 
aligned” with HBase design and should give all the benefits (less boilerplate 
in the code, column filters etc).

Does anyone use single-column design in production installations for 
performance or other reasons?

Thanks,
 Kirill

[1] https://blogs.apache.org/hbase/entry/the_effect_of_columnfamily_rowkey 
<https://blogs.apache.org/hbase/entry/the_effect_of_columnfamily_rowkey>, see 
Case #5

Reply via email to