Raid 5 Vs Raid 0

mir

Limp Gawd
Joined
Apr 4, 2007
Messages
200
In terms of read performance.
I have a option to use 3 WD 500 Gig Black HDD.
I really need 1TB so I was thinking about building a raid 5.
Since I have MB built in Intel ICH9R which going to do Raid 5.
Which got me thinking this could be very slow due to one pairty bit.
I know for write it would be slow hower in terms of read how it would be ?
 
RAID5 will be slightly slower than RAID0, but probably not by much. However the Intel implementation isn't particularly good and RAID5 read performance may suffer.

Just set it up and run some benchmarks, shouldn't take long to test out...
 
ICH9R works well enough for 1 or 0 but I would not throw 5 at it. 5 works infinitely better when your controller has some cache.
 
For having a cache controller I got a have some cash. :D
I think I will throw at it and see how well or badly it works
 
RAID-5 using onboard RAID controllers like the ICHxxR series, NVRAID controllers integrated into nForce chipsets and of course anything from JMicron will be relatively slow. RAID-5 can perform well but it has a ton of overhead. This overhead really needs a XOR processing engine like Intel's IOP333 chip in order to perform well. Hybrid/software RAID-5 just usually isn't all that great. I've heard people say that it's great on Linux based systems but my one experience with that says otherwise. In any case RAID0 and RAID1 are fine for onboard RAID controllers because there isn't much in the way of processing overhead necessary to make those RAID levels work.

Intel's onboard RAID implementation is as good as anyone's but it really isn't all that good compared to the controllers out there. For the price you can't beat it but if you are serious about RAID-5 go with a real RAID controller if you can afford it.
 
Sure you can, Linux softraid :p

Umm. No. I prefer to use RAID arrays with a relevant operating system. Linux doesn't qualify. (I hate to say it but it's true.) In any case ICHxxR / Matrix RAID or NVRAID / JMicron solutions come with just about every motherboard sold today. If you have a computer, most likely it's RAID capable.
 
Umm. No. I prefer to use RAID arrays with a relevant operating system. Linux doesn't qualify.

I prefer to use RAID on a server, where Linux is a more relevant operating system than Windows. Running RAID on a desktop is silly and 'not relevant'.
 
I prefer to use RAID on a server, where Linux is a more relevant operating system than Windows. Running RAID on a desktop is silly and 'not relevant'.

QFT

I can't see the point of running RAID 5 on a workstation. Keep things separate or you're asking for trouble.
 
Are those raid class drives? If not, you might see issues.

Also, using onboard raid controller will get you full of lols
 
I prefer to use RAID on a server, where Linux is a more relevant operating system than Windows. Running RAID on a desktop is silly and 'not relevant'.

Linux is certainly relevant in the server market but to say it's more relevant than Windows is naive. To say RAID isn't relevant or that running RAID on a desktop or workstation is silly is amazingly short sighted. There are plenty of reasons for running RAID on a desktop / workstation. I will agree that it is somewhat unusual and not generally worthwhile but again, in certain circumstances I think it makes a whole lot of sense.
 
Linux is certainly relevant in the server market but to say it's more relevant than Windows is naive.

Your joking right? Lets take a simple example from netcraft:
Code:
Developer	August 2009	Percent	September 2009	Percent	Change
Apache	104,611,555	46.30%	105,416,925	46.62%	0.33
Microsoft	49,579,507	21.94%	49,615,010	21.94%	0.00
qq.com	30,278,988	13.40%	30,069,048	13.30%	-0.10
Google	14,213,976	6.29%	13,767,338	6.09%	-0.20
nginx	11,502,109	5.09%	12,676,238	5.61%	0.52

Or how about from the super computer world, top500.org

Linux 77.40 %
Windows HPC 2008 1.00 %

I started out fighting hard for Microsoft, I've now migrated hundreds of systems/servers to Linux to deliver the uptime required.
 
Your joking right? Lets take a simple example from netcraft:
Code:
Developer	August 2009	Percent	September 2009	Percent	Change
Apache	104,611,555	46.30%	105,416,925	46.62%	0.33
Microsoft	49,579,507	21.94%	49,615,010	21.94%	0.00
qq.com	30,278,988	13.40%	30,069,048	13.30%	-0.10
Google	14,213,976	6.29%	13,767,338	6.09%	-0.20
nginx	11,502,109	5.09%	12,676,238	5.61%	0.52

Or how about from the super computer world, top500.org

Linux 77.40 %
Windows HPC 2008 1.00 %

I started out fighting hard for Microsoft, I've now migrated hundreds of systems/servers to Linux to deliver the uptime required.

I don't know the actual total installation base numbers of each platform. Yes I am aware that on the super computer front and in regard to mainframes and things of that nature that Windows is virtually non-existant in that marketplace. I disagree with any assertion that Linux is more relevant than Windows. As a contractor I've worked at more sites and for more companies that I can easily recount. Almost all of them had more Windows servers than Linux or Unix servers. In fact one thing I found strange is that the smaller companies often have zero Linux or Unix servers and rely purely on Windows for their infrastructure. Unlike you I've actually migrated several servers from Unix/Linux to Windows Server based OSes.

Here is an interesting read: http://en.wikipedia.org/wiki/Comparison_of_Windows_and_Linux Of course it's in dispute according to Wikipedia, but it's interesting. There it estimates the total installation base of systems as being a bit different than you do. Of course the Linux vs. Windows desktop marketshare numbers are of course fairly accurate. The supercomputer numbers I agree with, and the Linux vs. Windows server numbers are something I know little about.

This article here would also disagree with the notion that Linux is more relevant than Windows: http://www.linux-watch.com/news/NS5369154346.html It doesn't say that specifically, no. It does state that Linux only has 12.7% of the server market share.

This article is interesting as well: http://www.linuxinsider.com/story/48999.html?wlc=1254072252 In this one they point out that Windows has supplanted Unix as the dominant server operating system. They have also noted that Linux growth is increasing as well but it's taking away more of Unix's market share than anything.

On another note: As a contractor I've worked for companies of varying sizes and oddly enough the smaller companies I've worked with usually have almost zero Linux servers. I also happen to know that the State of Texas Datacenter has more Windows servers than Linux based servers.

Here is yet another example: http://www.eweek.com/c/a/Linux-and-Open-Source/Linux-Losing-Market-Share-to-Windows-Server/

Just using a Google search I found several pages of similar information. So I'm going to go ahead and say that Windows is probably still more "relevant" than Linux is. I'm not bashing Linux in any way. From what I've seen and understand, Windows is just dominant in the market and (to be fair, it's all I work with anymore. I don't typically work with Linux servers at all.) therefore, I think my point still stands. Also I find Linux to be slightly less relevant if for no other reason than the fact that every IT professional out there is probably familiar with Windows and there are far fewer that are famliar with Linux/Unix. I've known Linux/Unix proponents and server administrators that learned Windows later on and had no trouble doing so. While many Windows administrators looking to expand their skill set struggle to learn Linux.

Back to the original topic I wonder how many Linux/Unix systems use software RAID vs. hardware RAID solutions? Perhaps you can shed some light on this subject. Most of the Linux/Unix servers I've seen use hardware RAID. I knew one guy who turned off all the RAID controllers on a Linux server and used soft RAID to create the array. In that one experience that I had (I supported the Linux server after that guy left the company) I found that the soft RAID was problematic. It was always having to rebuild the arrays while our servers with hardware RAID controllers never had to unless a drive failed.
 
Raid 5 off Embeded ICH9R (No memory on controller)
Raid53WDBlack500ICH9R.jpg

Now Raid 0
Raid50WDBlack500ICH9R.jpg

HDD are WD 500 Blacks.
I think I am gonna go for Raid 0; cause perfomance loss little over 66%
But I do think built in controller card did pretty good without any ram on board or even it got nothing like XOR processing engine ( Intel's IOP333 chip)
What I want is as follow
4xWD1001FALS-RAID0.png
 
Back
Top