New Posts  All Forums:Forum Nav:

Which is faster?? - Page 4

post #31 of 43
Quote:
Originally Posted by the_beast View Post
Effectively RAID0 is a single 'drive'. But as there is twice as much hardware involved, and a failure from any part means the 'drive' fails, this means a RAID0 array is (effectively) twice as likely to fail as a single drive.

Having a single drive is better than 2 in RAID0 from a data safety view, but is nowhere near as safe as having your 2 drives either backing each other up or in a mirrored array (which uses the same number of physical disks, and so is therefore comparable).
This is the part I can't get my head around. In any array, each drive is independent and simply acts upon requests from the controller, therefore the chance of any one drive failing is not related to any other drive in the array (unless the chance of failure is related to vibration from the other drives, but that's another topic by itself).

Therefore to my mind, the chance of array failure in RAID 0 is equal to the chance of any one drive failing. However, the chance of any one drive failing is a product of wear & tear, damage from handling and manufacturing defects, and (in the case of wear & tear) will be variable over time.

However, surely the same factors will apply to a single non-arrayed drive?
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
post #32 of 43
Quote:
Originally Posted by parityboy View Post
This is the part I can't get my head around. In any array, each drive is independent and simply acts upon requests from the controller, therefore the chance of any one drive failing is not related to any other drive in the array (unless the chance of failure is related to vibration from the other drives, but that's another topic by itself).

Therefore to my mind, the chance of array failure in RAID 0 is equal to the chance of any one drive failing. However, the chance of any one drive failing is a product of wear & tear, damage from handling and manufacturing defects, and (in the case of wear & tear) will be variable over time.

However, surely the same factors will apply to a single non-arrayed drive?
Scenario:
5% of drives die after a year (1 drive in 20)

Create a 3 disk RAID 0 array from these drives.
Each drive has a 5% chance of failure. (1/20 + 1/20 + 1/20)
In RAID 0, if one drive dies, your whole array is toast.

Therefore, the risk to the array in the first year is the sum of the statistical probability of all drives, being 3/20, or 15%
Waiting on X399
(13 items)
 
  
CPUMotherboardGraphicsRAM
AMD Phenom II B57 @ X4 3.9 Gigabyte 790FXTA-UD5 Sapphire Radeon 290 8 GB G.Skill 2133 
Hard DriveCoolingOSKeyboard
250 GB 840 EVO Noctua NH-D14 Windows 10 Logitech K350 
PowerCaseMouseMouse Pad
Seasonic x750 Corsair 600T Logitech G100s Razer Goliathus Speed 
Audio
Plantronics Gamecom 788 
  hide details  
Reply
Waiting on X399
(13 items)
 
  
CPUMotherboardGraphicsRAM
AMD Phenom II B57 @ X4 3.9 Gigabyte 790FXTA-UD5 Sapphire Radeon 290 8 GB G.Skill 2133 
Hard DriveCoolingOSKeyboard
250 GB 840 EVO Noctua NH-D14 Windows 10 Logitech K350 
PowerCaseMouseMouse Pad
Seasonic x750 Corsair 600T Logitech G100s Razer Goliathus Speed 
Audio
Plantronics Gamecom 788 
  hide details  
Reply
post #33 of 43
Quote:
Originally Posted by beers View Post
Therefore, the risk to the array in the first year is the sum of the statistical probability of all drives, being 3/20, or 15%
Ahhh I see, so in statistics the chance of array failure is cumulative? For RAID 0 I've always seen it as concurrent (if that's the correct word to use in statistics - it's not a subject I've touched on), since all of the drives are operational at the same time, so the "count down to failure" would also be concurrent.
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
post #34 of 43
Say the chance of a drive failing is completely independent of any other action (which is a reasonable assumption, as just because one of your drives fails or otherwise, it does not mean any other drives you own will fail, or that they will keep working). For maths sake, lets call the failure rate 5%.

Now if you have 1 drive, the chance of it failing is (obviously) 5%.

If you have 2 independent drives, the chance that 1 of them fails is 5%. The chance that the other fails is 5%. If you assume that you have half of your data on each of them, that means you have a 5% chance of losing half your data, a 5% chance of losing the other half of your data, and a 0.25% chance of losing everything (5% x 5% = 0.25%, probability of 2 independent events occurring).

If you have a RAID0 array, then the chance of you losing everything is now dependent on either disk failing. Now the disk failure rate is still independent, so:

the chance of loss is = (chance of disk 1 failing AND disk 2 not failing) + (chance of disk 1 not failing AND disk 2 failing) + (chance of disk 1 AND disk 2 failing)

= (5% x 95%) + (95% x 5%) + (5% x 5%)
= 4.75% + 4.75% + 0.25%
= 9.75%
Or roughly 2 x 5% (you can add the probabilities with reasonable accuracy if the number of independent events is small and the probability of each event is small).

A more elegant way of working out the failure rate of a RAID0 array is to think of the failure like this:

Chance of data loss = 100% - (chance that nothing fails)
= 100% - (95%)^2
= 100% - 90.25%
= 9.75% as before

You can substitute the power of 2 in the above for the number of drives in the array, so for a 5% disk failure rate, the chance of an 8-drive array failing is given by:

= 100% - (95%)^8

= 100% - 66%

= 34% (note that now the number of events has increased, it's getting less accurate to just add the probabilities together. The chances of failure for a 1-8 drive array with the failure rates above are: 5%, 10%, 14%, 19%, 23%, 26%, 30%, 34%, although even for an 8 drive array the simple calculation works quite well, as 34% ~ 40%).
post #35 of 43
Quote:
Originally Posted by parityboy View Post
Ahhh I see, so in statistics the chance of array failure is cumulative? For RAID 0 I've always seen it as concurrent (if that's the correct word to use in statistics - it's not a subject I've touched on), since all of the drives are operational at the same time, so the "count down to failure" would also be concurrent.
the chance of each drive failing doesn't depend on the other drives - it is effectively a random event, so you don't really have a 'countdown'. Say for example your drives have an average life of 5 years - this means one drive might last for 10 years, but that doesn't matter if the other drive in the array failed after 5 minutes.
post #36 of 43
Cheers for the explanation, I think I see things a little better now. I think we're saying the same thing, but coming at it from a different angle.

Quote:
the chance of each drive failing doesn't depend on the other drives - it is effectively a random event, so you don't really have a 'countdown'.
It's not that random. Every drive wears, and if there's a defect in it from manufacture, or there's damage due to mishandling, it will simply fail more quickly. It might look random, because most times we can't tell that the drive is dying, but to my mind it isn't.

That's why I said that the countdown to failure would be concurrent - each drive's life span is independent of the others and could be longer or shorter, even with drives from the same batch.

The array fails at the moment when the drive with the shortest lifespan dies, so if a single drive has a 50% chance of failing in the first <period of time>, then the chance of catastrophic array failure would be 50%, or at least a minimum of 50%.

I understand that you also have to take the probability of the other drives failing into account, though. Not that it'd matter much with those odds.
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
post #37 of 43
This does not help with the 10k discussion BUT you can see the difference in platter performance and SSD performance. This does not touch on the obvious access time difference which is one of many reasons SSD drives are great.

Comparing a 4 platter disk array to a 2 SSD array. Obviously the 4 disk array increases the chance of failure.

Tool is ATTO

4x160GB 7200rpm 16mb sataII drives in RAID 0



2x64B C300 SSD drives in RAID 0
(64GB C300 SSDs only offer 75mb write, the 128's will offer double that)




Edited by Dark - 1/19/11 at 10:48am
un-overkill
(16 items)
 
Hadron 290x
(10 items)
 
 
CPUMotherboardGraphicsGraphics
Xeon E5-2695v3 Asus X99 Workstation Titan X Pascal Titan X Pascal 
RAMHard DriveHard DriveHard Drive
128GB Samsung 2133mhz ECC Samsung 950 Pro 512GB 3x Samsung 1TB SSD 1x 1.92TB SSD 
CoolingOSMonitorKeyboard
Corsair H105 Win10 x64 Acer Predator X34 Logitech G710+ 
PowerCaseMouseAudio
LEPA G1600 Corsair Air 540 Logitech G700S Bose 
CPUMotherboardGraphicsRAM
i5-4670k ASRock Z87E-ITX R9 290x 8GB Dominator 2Ghz CL9 
Hard DriveHard DriveCoolingOS
Samsung 250GB SSD Seagate 3TB  EVGA ACX ITX Win8.1 ENT 64 
MonitorCase
Dell U2412H EVGA Hadron Air 
  hide details  
Reply
un-overkill
(16 items)
 
Hadron 290x
(10 items)
 
 
CPUMotherboardGraphicsGraphics
Xeon E5-2695v3 Asus X99 Workstation Titan X Pascal Titan X Pascal 
RAMHard DriveHard DriveHard Drive
128GB Samsung 2133mhz ECC Samsung 950 Pro 512GB 3x Samsung 1TB SSD 1x 1.92TB SSD 
CoolingOSMonitorKeyboard
Corsair H105 Win10 x64 Acer Predator X34 Logitech G710+ 
PowerCaseMouseAudio
LEPA G1600 Corsair Air 540 Logitech G700S Bose 
CPUMotherboardGraphicsRAM
i5-4670k ASRock Z87E-ITX R9 290x 8GB Dominator 2Ghz CL9 
Hard DriveHard DriveCoolingOS
Samsung 250GB SSD Seagate 3TB  EVGA ACX ITX Win8.1 ENT 64 
MonitorCase
Dell U2412H EVGA Hadron Air 
  hide details  
Reply
post #38 of 43
http://www.newegg.com/Product/Produc...me=15000%20RPM

Get one of those. 0 RPM > 15,000 RPM > 10,000 RPM > 7,200 RPM
Black Sand
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.8GHz ASUS Maximus IV Extreme-Z EVGA GTX 580 SLI 8GB Corsair Dominator GT @ 1885 MHz 
Hard DriveOSMonitorPower
60GB Mushkin Callisto Deluxe +1TB WD Caviar Black Windows 7 Home Premium ASUS VW246 Corsair HX-850 
Case
HAF 932 
  hide details  
Reply
Black Sand
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.8GHz ASUS Maximus IV Extreme-Z EVGA GTX 580 SLI 8GB Corsair Dominator GT @ 1885 MHz 
Hard DriveOSMonitorPower
60GB Mushkin Callisto Deluxe +1TB WD Caviar Black Windows 7 Home Premium ASUS VW246 Corsair HX-850 
Case
HAF 932 
  hide details  
Reply
post #39 of 43
Can you put two SSDs in an array? Or does that make the universe explode?
First Build
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q9450 @ 3.2GHz Asus Rampage Formula evga GTX 580 (2x2GB) OCZ-Reaper 1066 
Hard DriveOSMonitorPower
VRaptor 150GB, 640GB WD Vista Ultimate 64-bit SP1 Samsung 22" 2232BW Thermaltake 850W 
  hide details  
Reply
First Build
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q9450 @ 3.2GHz Asus Rampage Formula evga GTX 580 (2x2GB) OCZ-Reaper 1066 
Hard DriveOSMonitorPower
VRaptor 150GB, 640GB WD Vista Ultimate 64-bit SP1 Samsung 22" 2232BW Thermaltake 850W 
  hide details  
Reply
post #40 of 43
Quote:
Originally Posted by mst3k View Post
Can you put two SSDs in an array? Or does that make the universe explode?
You can put as many as you can fit on the mobo. Over 2 SSD's and your computer starts to divide by zero though.


Edited by blackbalt89 - 1/19/11 at 12:18pm
Black Sand
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.8GHz ASUS Maximus IV Extreme-Z EVGA GTX 580 SLI 8GB Corsair Dominator GT @ 1885 MHz 
Hard DriveOSMonitorPower
60GB Mushkin Callisto Deluxe +1TB WD Caviar Black Windows 7 Home Premium ASUS VW246 Corsair HX-850 
Case
HAF 932 
  hide details  
Reply
Black Sand
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.8GHz ASUS Maximus IV Extreme-Z EVGA GTX 580 SLI 8GB Corsair Dominator GT @ 1885 MHz 
Hard DriveOSMonitorPower
60GB Mushkin Callisto Deluxe +1TB WD Caviar Black Windows 7 Home Premium ASUS VW246 Corsair HX-850 
Case
HAF 932 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: RAID Controllers and Software