RAID 0 setup: What do I need?

[grunt_408]

I currently have a 300 Gb Hdd How much faster would my pc be with a second 300 Gb.?

 

[IggSter]

RAID0 gives up to 40% performance increase on synthetic tests only. In applications and games it will only increase performance by 2% in some cases and you can find that performance can decrease by up to 2% in some cases as well from the overhead from the RAID controller.

If you want proof, go here: http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=1

Also, even if you use a very good external RAID controller, there is very little difference in the outcome except the overhead might be an unoticable fraction less.

I'm sorry but but I have to correct a few things here:

#1 the anandtech article is dated July 1st, 2004 so those figures are completely useless now.
#2 the system specs are pityful for testing a raid setup (again due to the age of test). There would be so many bottlenecks in that system.
#3 the testing method fo games is totally flawed. Loading a level in a game is hardly a comprehensive real life test.

 

[IggSter]

I currently have a 300 Gb Hdd How much faster would my pc be with a second 300 Gb.?

There is no hard and fast answer to this unfortunately as it depends on so many other factors but i will try and give you a general feeling for what to expect:

If you have a raid array and you run any form off application that thrashes the disk (BF2 is one that springs to mind also any database/video editing-encoding apps) you will notice your PC will feel far more responsive in general.

If you move a lot of large chunks of data around (100's of MB or more) copy times will be greatly reduced (if copying from raid to raid)

Large application load time will be reduced by a noticable amount especially so if application consists of large files (the smaller the files the less the benefit)

Defragmentation of drive becomes much less of an issue ie, your drives will need defragmenting less often and the slow down due to defragmentation is much less noticable.

Here is a couple of real life examples from one of my PCs

PC has C: which is a normal 160gig sata2 drive and R: which is a 160gig raid0 array (2 x 80gig drives)

Please note this is a working PC and was last given a fresh install of XP around 18mths ago.

Load time of a game called Eve:

c: 12sec

r: 8sec

Copy time of an acronis 3.8gig disk image:

r:\folder1 > r:\folder2 (this will test combined read/write of raid) - 2m 34sec

c:\folder1 > c:\folder2 (this will test combined read/write of sata2) - 3m 1sec

Nice improvement there I think you will agree.

r: > c: (tests read from raid and write to sata2) - 1m 12sec

c: > r: (tests read from sata2 and write to raid) -1m 35sec

This shows that reads from raid are faster than using standard sata2 drive and im sure that if I had 2 raid arrays available on same PC copying from one to the other would produce a figure of well below 1m.

I hope that helps.

 

[Hitsugaya_Toushirou]

There is no hard and fast answer to this unfortunately as it depends on so many other factors but i will try and give you a general feeling for what to expect:

If you have a raid array and you run any form off application that thrashes the disk (BF2 is one that springs to mind also any database/video editing-encoding apps) you will notice your PC will feel far more responsive in general.

If you move a lot of large chunks of data around (100's of MB or more) copy times will be greatly reduced (if copying from raid to raid)

Large application load time will be reduced by a noticable amount especially so if application consists of large files (the smaller the files the less the benefit)

Defragmentation of drive becomes much less of an issue ie, your drives will need defragmenting less often and the slow down due to defragmentation is much less noticable.

Here is a couple of real life examples from one of my PCs

PC has C: which is a normal 160gig sata2 drive and R: which is a 160gig raid0 array (2 x 80gig drives)

Please note this is a working PC and was last given a fresh install of XP around 18mths ago.

Load time of a game called Eve:

c: 12sec

r: 8sec

Copy time of an acronis 3.8gig disk image:

r:\folder1 > r:\folder2 (this will test combined read/write of raid) - 2m 34sec

c:\folder1 > c:\folder2 (this will test combined read/write of sata2) - 3m 1sec

Nice improvement there I think you will agree.

r: > c: (tests read from raid and write to sata2) - 1m 12sec

c: > r: (tests read from sata2 and write to raid) -1m 35sec

This shows that reads from raid are faster than using standard sata2 drive and im sure that if I had 2 raid arrays available on same PC copying from one to the other would produce a figure of well below 1m.

I hope that helps.

You didn't mention the HHDs used. The 80GB HHD could be originally higher performance than the 160GB, which means the results difference is enlarged unfairly. Well, no doubt transferring data between the RAID system and a separate source/destination will get a reasonable boost in timings but you need a more controlled comparison since these results could be exaggerated.

 

[Mussels]

Gigabit network goes up to 125MB/s but using RAID won't necessarily allow you to use all of it so you won't get all of a speed increase in network transfers but it'll be more than applications and games i guess but with a decrease in reliability.

Just from my personal testing i found that even average throughput operations goes by this rule i made (assuming you have no other bottlenecks):
Slowest HHD Bandwidth x (1+0.20^(1/(Number of HHD - 1))
So even with 5 HHD with 60MB/s bandwidth then even pure throughput operations like transferring over the network will only give about 100MB/s but at the cost of a massive reduction in reliability of your drives and near 0% performance increase in applications and games.

Though those are my personal tests so you don't have to take if you don't want to.

no reliability loss due to raid 5 - not raid 0. raid 5 enahcnes the reliability.
I am NOT using it for games - its for storage and backups, and moving 1TB of data takes a shitload of time when upgrading/installing more drives internally.

 

[IggSter]

Disks are:

Samsung HD160JJ with Noiseguard disabled (fairly typical Sata2 drive)

Raid drives: Hitachi Deskstar 7K160 80GB SATA-II 8MB x 2 (again fairly typical drives)

And with regard to Gigabit Ethernet, it is usual to run this in full duplex mode and thus get a max sustained throughput of 250MB/sec (this depends entirely on the NIC/Switch/Router)

 

[Mussels]

Disks are:

Samsung HD160JJ with Noiseguard disabled (fairly typical Sata2 drive)

Raid drives: Hitachi Deskstar 7K160 80GB SATA-II 8MB x 2 (again fairly typical drives)

And with regard to Gigabit Ethernet, it is usual to run this in full duplex mode and thus get a max sustained throughput of 250MB/sec (this depends entirely on the NIC/Switch/Router)

full duplex means absolutely nothing. you cannot get faster than 125MB/s.

Full duplex means it can do full speed both ways at once, which has always been a bukllshit advertising scam - you CANT get full speed both ways unless you're using a lot of hard drives on each end (seperate ones for each transfer) and its both ways at once - its 2x125 (different directions) NOT 1x250MB/s (double speed transfers)

Anyone who argues that point, its like saying a quad core is twice as fast as a dual core.

 

[IggSter]

That is partly true but:

1. full duplex disables collision detection on an ethernet interface, therefore a network interface never has to listen for collisions or retry = greater throughput.

2. full duplex allows a network interface to send and receive at the same time, and for every packet you send your will get an ACK (as minimum) returned. Therfore even if you are only sending a file, you will be receiving a shed load of data back = greater throughput.

So in essence in half duplex you have 125MB avail to send and receive, while in full duplex you have 125MB to send AND 125MB to receive (think along the lines of a bridge across a river with only one lane vs a bridge with 2 lanes and there is traffic crossing both ways. The two lane bridge will be much faster.

 

[surfsk8snow.jah]

Anyone who argues that point, its like saying a quad core is twice as fast as a dual core.

What, it's not??

 

[Mussels]

That is partly true but:

1. full duplex disables collision detection on an ethernet interface, therefore a network interface never has to listen for collisions or retry = greater throughput.

2. full duplex allows a network interface to send and receive at the same time, and for every packet you send your will get an ACK (as minimum) returned. Therfore even if you are only sending a file, you will be receiving a shed load of data back = greater throughput.

So in essence in half duplex you have 125MB avail to send and receive, while in full duplex you have 125MB to send AND 125MB to receive (think along the lines of a bridge across a river with only one lane vs a bridge with 2 lanes and there is traffic crossing both ways. The two lane bridge will be much faster.

yes. thats 2x125. NEVER can you reach 250MB/s, as each direction is capped at 125 maximum.
You cant even get gigabit thats not full duplex, its been 5 years or more since ive seen a half duplex network switch for sale.
half duplex is LESS than 125, full is 125. you mentioned 250, and i'm just making sure no one gets that impression.