General RAID question

I understand the difference between the various RAID modes. What I'm trying
to find out is how big the performance hit is in using RAID 0 to mirror
data. Are we talking a few ms here and there or many seconds?

I guess what I really want to know is if it is a noticeable delay to the end
user.

Thanx,

Vic Baron

--
There are 10 kinds of people - those who understand binary and those who
don't

hardware Raid has no such penalty. Software mirroring on the other hand
uses the CPU to perform the work.

Vic Baron wrote:

I understand the difference between the various RAID modes. What I'm trying
to find out is how big the performance hit is in using RAID 0 to mirror
data. Are we talking a few ms here and there or many seconds?

I guess what I really want to know is if it is a noticeable delay to the end
user.

Thanx,

Vic Baron

Using hardware raid - no noticable performance hit

"Vic Baron" wrote in message
...
I understand the difference between the various RAID modes. What I'm
trying
to find out is how big the performance hit is in using RAID 0 to mirror
data. Are we talking a few ms here and there or many seconds?

I guess what I really want to know is if it is a noticeable delay to the
end
user.

Thanx,

Vic Baron

--
There are 10 kinds of people - those who understand binary and those who
don't

raid 0 won't give a performance HIT, it will IMPROVE performance in most
disk intensive operations: booting/launching, defragging, searching,
virus/spyware scanning, backup, large file conversions. the boost in
other "everyday" things is not noticable (Word won't type faster!)

most raid 0 solutions today are built-in to the main chipset, and use a
combination of software (in the driver) and some hardware assist (in the
chipset); for the most part the "overhead" is more than made-up-for by the
performance boost.

but remember, if one drive in the array fails then you loose it all. so
backup backup backup (good advice for non-raid 0 too!).

wrote in message
...
raid 0 won't give a performance HIT, it will IMPROVE performance in most
disk intensive operations: booting/launching, defragging, searching,
virus/spyware scanning, backup, large file conversions. the boost in
other "everyday" things is not noticable (Word won't type faster!)

most raid 0 solutions today are built-in to the main chipset, and use a
combination of software (in the driver) and some hardware assist (in the
chipset); for the most part the "overhead" is more than made-up-for by the
performance boost.

but remember, if one drive in the array fails then you loose it all. so
backup backup backup (good advice for non-raid 0 too!).

I think the OP screwed up a bit here. He referred to RAID0 as mirroring.
RAID0 is *striping*, RAID1 is *mirroring*. I assume the OP actually meant
mirroring.

The whole purpose of RAID0 is to increase performance, so everything stated
in this post above is correct, including the cautionary notes, for striping.

However, since I believe the OP was actually referring to mirroring, the
situation is slightly different. The performance hit for hardware mirroring
is negligible. There's also so much file buffering taking place and delayed
writes, it would be very difficult to even measure the differences unless
you disabled these functions (which most software performance tools would
do, such as SiSoftware Sandra). In fact, your read performance may increase
slightly since it's theoretically possible for the raid controller to read
BOTH HDs in parallel to retrieve a file. Most people use mirroring to
protect data (as opposed to OS files), which is typically read far more
often than written. So for most cases, there just isn't much need to worry
about performance hits when it comes to hardware RAID.

Of course, if you're referring to software RAID, the situation may be far
different since there is no hardware assist.

Thanx, Jim - you are correct - brain cramp - I was referring to RAID 1 (
mirroring ). Currently running RAID 5 ( 3 - 9G drives/15G total). Daily tape
b/u. Probably going to upgrade the server and was looking at alternatives.
Of course, as the saying goes"If it ain't broke, don't fix it". The R5 is
working fine, probably should just leave it as is.

Thanx again all,

Vic

"Jim" wrote in message
...

wrote in message
...
raid 0 won't give a performance HIT, it will IMPROVE performance in most
disk intensive operations: booting/launching, defragging, searching,
virus/spyware scanning, backup, large file conversions. the boost in
other "everyday" things is not noticable (Word won't type faster!)

most raid 0 solutions today are built-in to the main chipset, and use a
combination of software (in the driver) and some hardware assist (in the
chipset); for the most part the "overhead" is more than made-up-for by
the
performance boost.

but remember, if one drive in the array fails then you loose it all. so
backup backup backup (good advice for non-raid 0 too!).

I think the OP screwed up a bit here. He referred to RAID0 as mirroring.
RAID0 is *striping*, RAID1 is *mirroring*. I assume the OP actually meant
mirroring.

The whole purpose of RAID0 is to increase performance, so everything
stated
in this post above is correct, including the cautionary notes, for
striping.

However, since I believe the OP was actually referring to mirroring, the
situation is slightly different. The performance hit for hardware
mirroring
is negligible. There's also so much file buffering taking place and
delayed
writes, it would be very difficult to even measure the differences unless
you disabled these functions (which most software performance tools would
do, such as SiSoftware Sandra). In fact, your read performance may
increase
slightly since it's theoretically possible for the raid controller to read
BOTH HDs in parallel to retrieve a file. Most people use mirroring to
protect data (as opposed to OS files), which is typically read far more
often than written. So for most cases, there just isn't much need to
worry
about performance hits when it comes to hardware RAID.

Of course, if you're referring to software RAID, the situation may be far
different since there is no hardware assist.