MySQL Lists are EOL. Please join:

List:General Discussion« Previous MessageNext Message »
From:beacker Date:December 24 2005 2:08am
Subject:Re: merge table: speed benefits?
View as plain text  
Tom Cunningham writes:
>(a) The docs say that spreading the underlying tables across different
>disks can make queries faster. I don't quite understand how this will
>work in a normal query: if I do a SUM(amount) over the entire table,
>will it be quicker if the table is spread across different disks? I
>don't see how mysql can avoid doing this sequentially.

     Multiple disks definitely helps, since in my case it allows me
to have N times the available read speed and N times the available
number iops (IOs/sec).  In my case I have a  61GB merge table that
is based upon 180 separate myisam tables.  This table contains 487M
records.  The kinds of queries I'm doing, I would not be able to
handle cleanly without merge tables on 4.1.15.

     Since trying to back this up would be a nightmare, I rely upon
a standby master, and 3 query slaves for this data.  This way things
work reasonably, and I have some hope of surviving some kind of
hardware failure.  This is just one small portion of the substantial
data set I'm responsible for over at Technorati.
                  Brad Eacker (beacker@stripped)
merge table: speed benefits?Tom Cunningham21 Dec
  • Re: merge table: speed benefits?Gleb Paharenko21 Dec
    • Re: merge table: speed benefits?Tom Cunningham21 Dec
  • Re: merge table: speed benefits?beacker24 Dec
    • Re: merge table: speed benefits?Tom Cunningham25 Dec