On 01/16/2015 05:57 PM, Kian-Tat Lim wrote:
> I thought at one point we were considering having one MySQL server
> instance (worker node) per spindle, rather than having to manage the
> data distribution internal to a MySQL instance.
Yes, we were considering that, but that's not optimal either, unless you
have multiple cores per spindle, but even then there is shared table
access (non-partitioned tables) and it could get complicated. We really
do want to have more boxes with fewer cores, if we do all our I/O
optimization right.
In some sense, if qserv is perfect as advertised, the random seeks will
be so few that the penalty of using RAID (less seeks available in the
system) will be minor. Unicorns could exist, too.
-Daniel
########################################################################
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
|