Re: [squid-users] squid with linux or freeBSD?

On Thu, May 01, 2003 at 03:12:01PM +0800, fooler wrote:
>
> hmmm... im not really satisfied with your answers to my questions and please
> dont excuse that i dont understand how raid level works...
>

You will continue to be dissatisfied. A squid mailing list is not the
place for a deep and meaningful discussion about raid levels. You
seem to have a little knowledge about raid and you are misapplying it,
in doing so you are giving the wrong advice to people who may know
less than you do. This site:

<http://www.pcguide.com/ref/hdd/perf/raid/levels/single.htm>

seems to give a balanced and clear view of the different raid levels
and how they work. Read, absorb, understand.
 
>
> no.. not necessarily 1/3 of your parity bits... it really depends how many
> stripes per disk you have... for example...
>

That drawing is confusing and wrong. You dont have a parity block per
stripe the parity is calculated for the same groups of sectors on each
disk (I suspect how big that groups is is implementation dependent but
I am not sure) - so, for example sectors 0-99 on D1 are the parity
sectors for sectors 0-99 on D2 and D2, sectors 100-199 on D2 are the
parity sectors for the same sectors on D0 and D3... and so on. A
stripe overlays this and is the minimum chunk of data that will be
read/written to the raid. Tuning the stripe size needs careful thought.
 
>
> the above has 3 disks with 4 stripes per disk.... if disk 1 (D1) is failed,
> you lost 2/4 of your parity blocks and 1/4 for disk 2 and 3 respectively...
>

No - your diagram is wrong as I said above. Check the site I mention,
they have some nice diagrams detailing different stripe sizes on
different raid types.
 
> anyway.. this is already out of topic for squid mailing list.. thanks for
> your time spending with me...
>

Well and truly off topic. I do apologise for that but RAID is such a
misunderstood thing I felt you had to be corrected.

-- 
Brett Lymn