Re: Disabling Write-Caching
On Tue, Apr 10, 2007 at 01:43:54AM -0400, Greg Folkert wrote:
> On Tue, 2007-04-10 at 15:39 +1000, CaT wrote:
> > On Tue, Apr 10, 2007 at 01:28:38AM -0400, Greg Folkert wrote:
> > DB monkies generally like turning it off as it has proven to cause data
> > loss in certain corner cases. Good DB monkies tend to be very paranoid
> > of their data.
>
> Do these DB monkies also force multi-million dollar system with 32GB
> caching controllers (with battery backup for the RAM) to turn off the
> cache?
Those tend to have HDs which properly report the status of the data (ie
when theys ay 'data is on the platter' the data actually /is/ in the
platter. No foolin'.
> I've seen it, forcing the $COMPANY to invest in even larger machines,
> with diminishing results. Turn cache back on, it is like a whole nuther
> 20 processors added. We are talking Multi-vpath stuff here.
That depends on wether the gains made by turning the cache on outweigh
the potential disadvantages of turning the cache on. A good DB monkey
will provide said company with a risk analysis and get them to sign off
on one or the other.
> Good DBMonkies also force the DB software to do a sync'd write through
> to the drives or logical drive in any case. Blah, no tthe place to
> discuss this vitriol.
sync writes don't mean much when the HD lies.
> Screwy. And if they are using PATA drives on critical data, stupid.
> Unless they are behind a SAN or something similar... which is going to
> have huge caching involved caching.
It all depends on the needs of the people involved and the size of their
budget.
--
"To the extent that we overreact, we proffer the terrorists the
greatest tribute."
- High Court Judge Michael Kirby
Reply to: