Indexes can generate vast amounts of random i/o. Because of rotational latency, random i/o can really slow you down, especially if you are using IDE or SATA disks because they can't do tagged queueing like SCSI disks can.
If you have the budget for it, I would consider getting some solid state disks. Because they have extremely low latency you will be able to get full i/o bandwidth on your reads. If you can't afford those, consider adding more disks to your RAID array so that you can spread the reads over more spindles, which will help performance. --- matt ryan <[EMAIL PROTECTED]> wrote: > [EMAIL PROTECTED] wrote: > > >You may want more indexes but you might be getting > killed because you already have too many. > > > >To test - try loading into a table without indexes > and see if it makes a difference. > > > >At the very least - check to see if the primary > index which starts with 'dic' can make your special > 'dic' index superfluous. > > > >If write speed is a bottleneck you might consider > Raid-1 instead of Raid-5. > > > >Reading lots of rows via index is a killer. > Depending on your hardware it may be cheaper to > table scan 50 rows than to read 1 via index. > However, this requires partitioning of the data > based on some column which appears in every query > and acts as an initial filter. If you are lucky > enough to be in that situation - consider a MERGE > table. > > > > > > > > > > These tables are merged, the total table size is > huge, on this > particular table, it's , 45,449,534 rows, however, > all the merge tables > combined are 258,840,305 records > > perhaps I should reorder the pimary key, putting the > longest most unique > record up front, and the least unique at the end, > would that speed up > the key check? I can tell that almost everything > is read IO, very > little write IO > > -- > MySQL General Mailing List > For list archives: http://lists.mysql.com/mysql > To unsubscribe: > http://lists.mysql.com/[EMAIL PROTECTED] > > -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
