Sign in with
Sign up | Sign in
Your question

RAID0 "realistic" performance benefit?

Tags:
  • Hard Drives
  • NAS / RAID
  • Performance
  • Storage
Last response: in Storage
Share
January 5, 2005 4:29:09 PM

Hi Everyone!

It is time for me to upgrade my system again, from an Athlon XP 2200+ to an Athlon 64 3000+ Winchester. I would also like to move from my slow PATA hard drive to a SATA. I plan on going with the MSI K8N Neo Platinum motherboard because it has all the bells and whistles, and seems to be highly prized here.

My question though was regarding RAID0. I am very familiar with server-based RAID, and have done extensive benchmarking on Ultra320 15k RPM drives running in all the different RAID modes, and I have been absolutely astonished with the speed boost with going RAID10 or RAID0 over even RAID5. A server running RAID0 next to an identical server running RAID5 is a night and day difference, not exaggerating, I have seen two identical servers, one in RAID5 and one in RAID0 next to each other, and the RAID0 server loads a service pack in 1/4th the time that the RAID5 machine takes.

I was assuming, perhaps falsely, that SATA RAID would provide similar speed benefits, so I was just about to order a pair of 250GB 7.2krpm/8mb Seagate SATA drives and run them in RAID0 on my new system until I read the following article over at Anandtech, showing an incredibly low “real-world” performance gained. I was horrified to see performance gains of 5% or lower over a single drive! Below is the link to that review:
http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=...

I was under the impression that these onboard RAID cards used their own RAID controller and did not rob the CPU to manage the RAID, but perhaps that is not the case?

The members of the Toms Hardware board have been infinitely helpful to me before, and I am hoping I can gain some clarity on this issue. Has anyone here gone from a single-drive situation to a RAID0 situation using the onboard RAID controller and what was your performance like? Has anyone gone from an onboard RAID to using a PCI card such as the value-priced Promise SATA card, did that give a large performance boost? And in general, will I be “wowed” at the speed of a SATA RAID0 solution over a single SATA drive?

I certainly appreciate any and all replies!

Thanks!

Eric

More about : raid0 realistic performance benefit

a b G Storage
January 6, 2005 12:35:01 PM

If you have to go with a RAID setup, you are probably best off with an external PCI card, like something from High Point. A legitimate RAID controller is going to set you back a lot closer to $100.
-Brett
January 6, 2005 4:19:54 PM

Most of those high end Ultra320 SCSI controllers are true hardware raids so they get a fairly decent performance boost. All onboard raid cards are software based for desktop machines. This does use some CPU time but with how fast cpus are getting it really doesn't make any sort of impact in performance. If you want a true hardware sata controller you are going to have to plunk out at least $250 or more. All 3ware controllers are true hardware controllers.
I get upwards of 90mb/s with my Raid0 36.7gb Raptors using the Sil3112 adaptor on my motherboard. I don't recall 100% what the transferspeeds were with a single drive but I think it was around 50-60 so the raid0 increase was decent.

<A HREF="http://www.folken.net/myrig.htm" target="_new">My precious...</A>
Related resources
January 6, 2005 4:20:51 PM

RAID0 improve continuous read/write performance. An OS drive spends 95% of it's time doing RANDOM reads. Hence the reason why improving seek times it more important than continuous read/write for an OS drive. Certain apps do require high continous read/write however, for those apps, RAID0 is a considerable benefit.

RAID0 doesn't really use any more CPU time than a standard hard disk controller. It just uses a modified write protocol to distribute the data differently. RAID5 consumes oodles of CPU time because of the XOR (parity) calculations required above the disk controller function.

<A HREF="http://www.anandtech.com/myanandtech.html?member=114979" target="_new">My PCs</A> :cool:
January 8, 2005 3:53:39 PM

Thank you very much for everyone's reply! I think at this time, since I am not a big gammer, nor do I really need a screaming fast PC, I will save the extra money and go with a single drive solution. And in the future get an external RAID card.

Thanks again!
January 9, 2005 4:53:15 PM

I like RAID0. I found that my system is just more snappy and fast. I do work with big video files and RAID0 is great for that. Problems with RAID 0 is that it create one big HDD, that at the end, can be slow. Just like your 2 x250 gigs HDD will create one big 500 gigs HDD. What you'll want to do is to create a small(?) 60 gigs partion, for example, for system, forcing the os to use the inner parts of the drive, thus, making it faster. Because, if you use all the space as one big partition, the files will be scatterred over all the drive killing the RAID0 performance advantage.



On my system, using the nvidia nforce3 sata raid controller with 2 seagate 160 gigs, the first partition, which is 60 gigs run at an average of between 100 and 110 MBytes/sec while the last one, a 5 gigs, for usual storage run at about 58 MBytes/sec. Using the whole disk will give you an average of about 65 MBytes of transfer, which is about the same as a fast standard HDD. On my system, I did 1 60 gigs partition, 2 100 gigs, one 40 gigs and 1 small 5gigs. I use the 2 first for the OS and work as video capturing and editing and the last ones for storage purpose, as they are a bit slower than the first ones. And give me the possibility to do a fresh install of windows without having to backup everything, I just move my data to another partition, reformat the 60 gigs partition, reinstall the os and programs and put my data back where it was before.

-Always put the blame on you first, then on the hardware !!!
January 9, 2005 9:50:35 PM

Greetings,

Here's a well written and empirical analysis
of RAIDs in various configurations:

http://www.storagereview.com/articles/200406/20040625TC...

This article may answer your question directly.

We recently assigned C: to a single WD 74GB Raptor
on an ASUS P4C800-E Deluxe, and never looked back.

It works marvelously fast with a 2.8GHz P4 512K L2 cache
and 800MHz FSB (Northwood core).

We got hit pretty hard by a virus last year
on an aging Windows 98/SE machine.

We now depend a LOT on Drive Image 7 to create and
restore "image" files of our C: partition on the
new ASUS motherboard with Windows XP/Pro.

This is the fastest way we know of recovering from
a destructive virus or worm.

This software (now acquired by Symantec and re-named
"GHOST") does not appear to work if C: is on a RAID 0,
however, and that is one of the main reasons why
we did not configure C: on a RAID of any kind.

We're planning right now to build an experimental
machine which will also have a single HD for C:,
plus a RAID 0 with 2 x SATA HDs @ 40GB each
(80GB total "striped"). We chose a single HD
for C: so we can easily restore C: from an
"image" file written by Drive Image 7.

On our special-purpose RAID 0, we plan to
store ONLY the Internet Explorer cache, and
possibly also the Windows swap file (for now).

Because the IE cache tends to get large,
the more so as we browse the Internet,
Drive Images of C: grow larger accordingly.

By moving the IE cache to different drives,
C: stays quite static.

Moreover, the Windows swap file is volatile and
does not need to be saved between shutdown
and startup. So, it too can be assigned
to such a RAID 0.

And, for our database, we will go with a
single large 300GB PATA Maxtor with 16MB cache
(which we just bought at Office Depot at 50% discount)
and possibly add future SATA drives of similar size.

This machine will plug into a Gigabit LAN, which
also has lots of storage space on the other nodes
for backup purposes, e.g. XCOPY /s/e/v/d (across
the Gigabit switch) of complex folders with 50,000+
discrete files.

Another way of insuring "snappy" program launch
speeds is to make sure you have extra RAM,
which reduces the need for swap file I/O
in the first place.

I hope this helps.


Sincerely yours,
/s/ Paul Andrew Mitchell
Webmaster, Supreme Law Library
!