FYI... see below. hmm, so our DC data sets have too many zeros, and we shouldn't be drawing too many conclusions regarding database file sizes on disk based on that data. Jacek On 02/21/2014 08:05 AM, Sergei Golubchik wrote:> > Try 'SHOW TABLE STATUS' in this database. You'll see that 50-column > table has a dynamic row format (because of your varbinary column), > while others have a fixed row format. > > That explains, basically, everything. In the fixed row format > there are no "record links" at all. Row length is calculated > as 4 bytes per int or float, 8 bytes per bigint or double, one > bit per NULL-able column, plus one bit as a "deleted row" flag. > Which matches exactly the MYD file > length, and the row length as shown by 'SHOW TABLE STATUS'. > > In the dynamic row format (see myisamchk -dvv small50), all numeric > columns got "no zeros" flag. So, if the value in the column is exactly > 0, it's not stored at all (but instead a special bit in the row header > is set). That's why a dynamic row is notably smaller in your > case - you apparently have lots of zeros in the table. > ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the QSERV-L list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1