Translating with DrWatson… this can take a few seconds the first time.
This is a genuine, complex translation. DrWatson protects commands, error codes, and log output while naturally translating the surrounding text. It’s translated once and saved.
Wang Jean asked whether raw devices still beat cooked files for a 300GB Informix data warehouse on Solaris. Thomas J. Girsch said cooked files can be fast but are less reliable because OS caching may hide unwritten data, notably at checkpoints. David Williams said raw still wins. Paul Watson advised raw, as its code path is always shorter.
Auto-generated by Claude from the posts below — may be imperfect; read the full thread.
Wang Jean — — source: Informix-list mailing list archive (1991-1998)
We have a 300GB data warehouse in Informix with Raw device in Sun Solaris.
Somebody suggested to use cooked files on faster disk drivers. I understand
that Raw device has better performance and more reliable. Is it still the
case? Are there any changes since Unix OS got lots of improvement now?
↪ replying to Wang Jean
Thomas J. Girsch — — source: Informix-list mailing list archive (1991-1998)
While it's true that sometimes cooked files can now be faster than raw
devices, they are generally considered to be somewhat less reliable. This
is because when you get that kind of speed out of cooked files, there's
usually some kind of RAM caching going on. Which means that Informix will
think that data has been written to disk when it really is still in RAM
waiting to write. This is particularly important at checkpoint time.
Some disk arrays also use cache, so you might experience this with raw
slices as well, but in these cases the cache is hardware-based (not
software-based) and is generally fault-tolerant.
HTH,
- Tom Girsch
TJG Technical Services
tgirsch@tjgtech.com
Wang Jean <WANGJ@amtrak.com> wrote in message
news:7uqluc$de7$1@news.xmission.com...
>
> We have a 300GB data warehouse in Informix with Raw device in Sun Solaris.
> Somebody suggested to use cooked files on faster disk drivers. I
understand
> that Raw device has better performance and more reliable. Is it still the
> case? Are there any changes since Unix OS got lots of improvement now?
>
↪ replying to Wang Jean
David Williams — — source: Informix-list mailing list archive (1991-1998)
In article <7uqluc$de7$1@news.xmission.com>, Wang Jean
<WANGJ@amtrak.com> writes
>
>We have a 300GB data warehouse in Informix with Raw device in Sun Solaris.
>Somebody suggested to use cooked files on faster disk drivers. I understand
>that Raw device has better performance and more reliable. Is it still the
>case? Are there any changes since Unix OS got lots of improvement now?
>
Stil the case.
--
David Williams
↪ replying to Wang Jean
Paul Watson — — source: Informix-list mailing list archive (1991-1998)
In article <7uqluc$de7$1@news.xmission.com>, WANGJ@amtrak.com (Wang Jean)
wrote:
>
> We have a 300GB data warehouse in Informix with Raw device in Sun
> Solaris.
> Somebody suggested to use cooked files on faster disk drivers. I
> understand
> that Raw device has better performance and more reliable. Is it still
> the
> case? Are there any changes since Unix OS got lots of improvement now?
>
>
Go for raw unless you have significant amounts of NVRAM in your RAID
controller and all the raid is being down at a hardware level. Even then
I'd still go for raw as I get my disk confirmations from the hardware
NVRAM. So as far as the OS is concerned I've got drives with
sub-millisecond response rates.
If you compare like with like then raw is always without exception
quicker, it has to be the code path to disk is significantly shorter.
Paul Watson # I don't suffer from
WF Software Ltd. # stress, I'm just
Tel. (+44) 1436 674729 # a carrier
Fax. (+44) 1436 678693 #
www.wfsoftware.com
Your privacy choices
We use strictly necessary cookies to make this site work. With your
consent we’d also use optional cookies for analytics and marketing. You can accept all,
reject all, or choose. Read our Cookie Policy.