I believe what you're saying is correct, James. The SMALL hint is calling scan.setSmall(true) on the scans for your query. There's some documentation on this here: https://issues.apache.org/jira/browse/HBASE-9488, but I'm having a hard time finding other documentation. Small scans prevent as many RPCs as would otherwise occur when a scan is executed.
Thanks, James On Wed, Jul 27, 2016 at 8:12 AM, Heather, James (ELS) < [email protected]> wrote: > On Tue, 2016-07-26 at 08:39 -0700, James Taylor wrote: > > - if you're querying for a single row, use the serial and small hints like > this: /*+ SERIAL SMALL */ > > > Quoting from the language reference: > > Generally, if the query is a point lookup or returns data that is likely > in a single data block (64 KB by default), performance may improve when > using this hint. > > Can you explain what's meant by "in a single data block"? Do you mean that > the result is located on a single region server, and in a single HDFS block > on that region server? > > In other words, if we're retrieving data from a single table, then it > ought to be a small number of rows with fairly closely located primary keys? > > James > > ------------------------------ > > Elsevier Limited. Registered Office: The Boulevard, Langford Lane, > Kidlington, Oxford, OX5 1GB, United Kingdom, Registration No. 1982084, > Registered in England and Wales. >
