Re: Raw Devices: Increased Performance?

In article <4quudo$eje_at_scel.sequent.com> krader_at_crg8.sequent.com (Kurtis D. Rader) writes:
>This makes me think of the favorite TLA of the sales force:
>
> FUD: Fear, Uncertainty, and Doubt
>
>joelga_at_rossinc.com (Joel Garry) writes:
>
>>There was a thread on this a while ago, with some folks claiming minimal
>>increase, while others claimed up to 300%.
 

>>Personally, I think the one time that someone unthinkingly writes over
>>the raw system will more than make up for any increase. In other words,
>>it requires greater control over administration than any place can be
>>expected to have over a long period of time.
>
>This is true only for raw disk partitions. If you are using a
>logical volume manager (ala Veritas VxVm or ptx/SVM) then the risk

OK, if you are. And ok if you only are. However, if you are using a mix of machines, some of which have it and some don't, that is even worse.

>of overwriting your raw database is no greater than overwriting a
>cooked database file. And in fact I have never taken a call from
>a customer who has accidentally destroyed his raw database. Yet

Well, I have.

>I have taken four calls (that I can remember) in the past six years
>from customers who have accidentally blown away their cooked
>database. For that matter, I ran both raw and cooked databases
>back when I was a sysadmin/DBA and never had a problem with either.

Good for you. Now how that applies to someone who would ask the question, "Should I use a raw partition?" is another question. They don't have as much experience as you. Noting your sequent.com address, I wonder how much DEC Unix volume manager experience you really have. I don't have any, so maybe you are right. My only experience with Veritas was on SCO, and it didn't work right. It did appeal to the perverse side of my sense of humor that the software that was supposed to increase reliability coredumpted. I don't assume any customer has an lvm (except AIX, and even that only from all the off-point flames, considering the original poster asked about DEC) and don't assume that it, as any software, is bug and user-error free.

>
>This is something I have studied extensively in my role designing
>system architectures for some of the largest Oracle sites in the

Let us hope that the largest Oracle sites in the world have some sophisticated and stable adminstration. Now how does that matter for some little place with a DEC box? From what I've seen, a lot of smaller shops say "You - Yeah, You. You get to learn unix and oracle now." Of course, my view is probably skewed, 'cause then they call me with "I haven't taken the courses yet, but we have this new computer..."

>world. Performance is actually pretty low on my list of reasons
>for favoring raw over cooked databases. A matrix I use when
>discussing this with clients shows the number of factors in favor
>of raw outweighing the number favoring cooked. That des not mean
>I always choose a raw implementation. It simply means that you
>shouldn't let FUD (see the start of the post) drive a decision.

Please share the matrix.

You didn't answer my point, which is that administration can deteriorate over time, and it can be worse with raw filesystems. You merely intimated that I am a salesperson or something.

>--
>Kurtis D. Rader, Senior Consultant krader_at_sequent.com (email)
>Sequent Computer Systems +1 503/578-3714 (voice) +65 223-5116 (fax)
>80 Robinson Road, #18-03 Currently on assignment in the
>Singapore, 0106 Asia-Pacific region

-- 
Joel Garry               joelga_at_rossinc.com               Compuserve 70661,1534
These are my opinions, not necessarily those of Ross Systems, Inc.   <> <>
%DCL-W-SOFTONEDGEDONTPUSH, Software On Edge - Don't Push.            \ V /
panic: ifree: freeing free inodes...                                   O