two HDs in RAID better than one large drive?

"Walter Mitty" wrote in message
...


They have the same rpm. Are you sure about that? Do the sectors hold
more on the outside (wider) section?


Same rpm = same time per revolution, no matter where along the radius you
are.

Which means either:
a) there is a greater data density in the centre (smaller revolution
circumference)

or

b) there are more sectors in the (longer) outer tracks, and all tracks have
a uniform (or in reality, similar) data density.

I know that all simplistic diagrams show each sector as being an arc of an
equal number of degrees [and indeed it used to be that way], but in actual
fact it varies - hence b) is the reality.

See
http://www.dewassoc.com/kbase/hard_d...structures.htm

Apparently, this technique is called Zone Density Recording (no, I didn't
know that either...)

"CJM" == CJM writes:

CJM There is a myth that RAID 0 is so much faster than a single
CJM drive, but it is patently not true.

There is no mythology, except that propogated by the ignorant. A
RAID 0 array will always be faster than the same drives not RAIDed.

CJM The fastest single drives
CJM (Raptors excluded) match RAID 0 in the majority of real world
CJM circumstances

They cannot match themselves when the same drives are RAIDed. Think.

CJM On the other hand, you face double the risk of failure!

Double of very small is still very small.
--
Freedom is not something that anybody can be given, freedom is
something people take.
~ James Baldwin

In comp.sys.ibm.pc.games.action Geek Dad wrote:
Are you joking? I hope so.

You do realise that PC users install the game onto their hard drive don't
you?

Right. And like I said, if anything happens to the hard drive, they
still have the original CD or DVD or their steam account to re-
download the game.

The OP is probably thinking of RAID 0 (striped) as opposed to RAID 1
(mirrored)

I agree mirroring for a gaming rig is a silly idea. The amount of data you
stand to lose is going to be relatively small, and easily handled by
traditional backup methods. It also results in absolutely zero
performance improvement.

However, striped can theorhetically deliever better performance because
you're splitting the read/write operations between 2 (or more) hard
drives. The downside here is that if you lose a hard drive, you lose
everything. Data loss isn't a big deal for a gaming rig, but I don't see
you getting enough of a noticable performance boost from this to be worth
the hassle and extra complexity. If anything, I'd rather have 1 smallish
hard drive (maybe 200-300GB) for the OS, and a separate larger drive
(500GB) for my games. Even when playing a game, you just aren't going to
be accessing the drive(s) enough to justify RAID.

RAID for servers makes sense because you want both redundancy (mirroring)
and better throughput (striping) which is why most use RAID5 or even
RAID10

--
It's not broken. It's...advanced.

"Bob Fry" wrote in message
...

There is no mythology, except that propogated by the ignorant. A
RAID 0 array will always be faster than the same drives not RAIDed.

As I've already explained, in real world scenarios, this is often not the
case. There will be benefits for transfer of *large* tranches of data, but
there is no benefit for lots of small pieces of data.



They cannot match themselves when the same drives are RAIDed. Think.

How about you think. Or just listen.

In most typical, real-world scenarios a single drive is just as fast as two
of said drives in RAID 0.


CJM On the other hand, you face double the risk of failure!

Double of very small is still very small.

The risk of my house burning down is also very small, but I still a) pay for
insurance, and b) make sure there are enough unobstructed exits.

The risk of failure *is* small, I've had 2 drives fail in the last year, One
was almost new and was a shock, but the other was perhaps 4 years old. I had
backups so I was OK, but if they had been raided, I would have lost twice as
much data. In my experience, the benefits are so minimal, the risks are real
and the impact of failure is potentially catastrophic.

In a gaming system, you might decide that it's a risk worth taking; but
unless the benefits were greater, I wouldn't bother. Having a new level load
in 8s rather than 9s is not sufficient to take such risks.

"CJM" == CJM writes:

CJM "Bob Fry" wrote in message
CJM ...
There is no mythology, except that propogated by the
ignorant. A RAID 0 array will always be faster than the same
drives not RAIDed.

CJM As I've already explained,

You haven't explained anything. You simply repeat the myths of the
uninformed. RAID 0 systems provide greater bandwidth than the same
drives unraided, in the lab, and in the real world. Latency is more
consistent across the drive also. The question for the OP is whether
that improvement makes any difference for games. I don't know since
I've never installed or used a game. If they don't use the disk much
after loading I wouldn't bother setting up RAID. If they do use the
disk during gaming, the OP will notice a performance improvement by
going to RAID 0.

CJM The risk of failure *is* small, I've had 2 drives fail in the
CJM last year, One was almost new and was a shock, but the other
CJM was perhaps 4 years old. I had backups so I was OK, but if
CJM they had been raided, I would have lost twice as much
CJM data. In my experience, the benefits are so minimal, the
CJM risks are real and the impact of failure is potentially
CJM catastrophic.

An external backup drive is a good idea for anyone with more than a
few hours worth of customization in their system, raided or not. So in
your example you did not lose anything, and would not have lost "twice
as much", because you had backups.

My only disk failure in years was the external enclosure of a backup
drive. I enjoy so much better performance with the RAID 0 setup that
I'll certainly keep it. In fact I've been moving most of my Program
Files and other directories over to it. At work all our new machines
will have a small boot disk and two large disks in RAID 0 with two or
three partitions.

--
Lying is done with words and also with silence.
~ Adrienne Rich

Doug Jacobs writes:

I agree mirroring for a gaming rig is a silly idea. The amount of data you
stand to lose is going to be relatively small, and easily handled by
traditional backup methods. It also results in absolutely zero
performance improvement.

Is this really the situation in Windows land? In Linux the software
RAID 1 provides faster reads, since obviously read operations can be
striped from a mirror as all disks in the array have the same data.

Then again, a gamer who develops mods or anything else might well want
a mirrored machine.

Data loss isn't a big deal for a gaming rig, but I don't see you
getting enough of a noticable performance boost from this to be
worth the hassle and extra complexity.

Yeah. As I recall, a few years ago both Storagereview and Anandtech
failed to show benefit for gaming from striping. One of the tested
games was Far Cry with its longish load times, but there was no help.
Apparently games don't do a large linear load into RAM, which is a
shame.

"Anssi Saari" wrote in message
...
Doug Jacobs writes:

I agree mirroring for a gaming rig is a silly idea. The amount of data
you
stand to lose is going to be relatively small, and easily handled by
traditional backup methods. It also results in absolutely zero
performance improvement.

Is this really the situation in Windows land? In Linux the software
RAID 1 provides faster reads, since obviously read operations can be
striped from a mirror as all disks in the array have the same data.


This is true, but in practice it makes very little difference.

And if take off the time Intel Matrix Manager spends rebuilding your array
because you sneezed at the wrong time or because the wind changed direction.
I have a 2 pairs of mirrored drives at work, and the one time a disk failed,
it took the array with it (ie. it corrupted the array before finally
failing).

Yeah. As I recall, a few years ago both Storagereview and Anandtech
failed to show benefit for gaming from striping. One of the tested
games was Far Cry with its longish load times, but there was no help.
Apparently games don't do a large linear load into RAM, which is a
shame.

This is the problem. In theory, there is a small % improvement, but only for
large files. In practice, any improvement is much smaller.