Raid users - your experiences

[BoDEAN]

Never did RAID setup before on my home pc. What are the advantages to going RAID? I assume you can use different brands/sizes for your raid configuration....What are real world results and benefits with running RAID over AHCI?

[Zanaffer]

I've used (software) RAID5 on my own system and (software) RAID0 and RAID1 at my job. I'll break it down.

RAID (in general) pros:

- Redundancy against HDD failure (except with RAID0)

RAID (in general) cons:

- RAID is not a substitute for backup, there are lots of things that can cause data loss/corruption that RAID can't help with which would be recoverable with a backup.

- Like yourself, there are lots of people who don't have experience with RAID. If you are not the one who performs the maintenance on your PC, I strongly suggest against RAID.

- The RAID levels with redundancy come with a penalty to usable space.

RAID0: Runs two (or more) drives as a single drive for the purposes of increasing speed.

RAID0 Pros

- Speed

RAID0 Cons

- No redundancy, none. If one drive in the array dies, you lose everything. It's basically computer AIDs (as there is no redundancy)

RAID1: Runs two drives as direct mirrors of each other, for the purposes of redundancy

RAID1 Pros:

- Possibly the best stability, if one of the HDDs dies you will still have all your data on the other.

- HDDs can be read on computers other than the one that the array was used in. This is useful if the raid controller/motherboard dies.

RAID1 Cons:

- Computer can/will still behave oddly and/or fail to boot if one of the drives dies and the RAID is the boot drive. This will persist until the failed drive is removed.

- You lose a whopping 50% of your storage space. (2x2TB Drives in mirror would give you 2TB of space total)

RAID5: Using 3 (or more) disks as a single volume, with "parity" information spread between drives

RAID5 Pros:

- Gives you redundancy against a single drive failure.

- Allows you to use more of your actual storage than RAID1. (Ex: 3x2TB HDDs gives 4TB of usable space, or 2/3)

Cons:

- Requires more disks than RAID1

- If the RAID controller or motherboard fails, it is difficult to access the data without another of the same type of controller.

- Drives are not readable on other PCs unless they use the same type of raid controller or without specialty software.

- Generating parity information requires CPU time, which results in slower write speeds.

I have since stopped using RAID5 on my home system. I was doing it using consumer grade drives, and the Seagate drives I was using did not like the combo of sleep + RAID and would go into an endless spinup/park cycle when the machine was woken from sleep. The short term solution was to disable sleep, but on my newest PC I opted to not use RAID for that reason.

[+Byt Subscriber¹]

In my experience, i tend to stay away from software raid.

if your board has a hardware raid built in then great, if not then get a little pci raid card.

I use RAID1 and its saved me many many times, the only issue for me is on my Asus board, you cannot set up alerts that alert you within Windows when a drive fails, so every now and again i have to go into the Nvidia Storage Manager to check the health of the RAID.

I have never had a PC not boot up correctly because 1 drive in a RAID1 Configuration has failed to be honest.

My current configuration is an SSD which has Windows 7 on it, and then two 1TB HDD's in RAID1. The thing perfoms brilliantly.

Always remember this though... RAID is not a form of backup! It is just to protect you against a hard drive failure. You should still backup your data to an external hard drive. I have friends that like to insist they have all their data backed up just because they have a RAID configuration :(

I was a fan of Cloud Backups at one stage, but when i had to download 100+GB from an online storage place like Mozy Backup, i was thinking to myself "Never again" just stick to external hard drives.

:)

[n_K]

Just a little correction, RAID1 the usable space is not 50%, it's the size of the smallest disk in the array, so 50% if you've got 2 drives, 33% if you've got 3 drives, 25% if you've got 4, etc.

There's a lot of different types of RAID levels, RAID6 for example extends RAID5 so that 2 concurrent disk failures can occur and the virtual drive will still work (If a drive fails whilst the virtual disk is being written to a backup drive, the virtual drive is dead.)

You also get nested RAIDs like RAID10/0+1/50/60 which is like like having a RAID5 for example mixed with RAID0, so you get the redundancy of RAID5 with the pros of much faster access of RAID0 but you need more disks to do it.

[+Fahim S. MVC]

In my experience, i tend to stay away from software raid.

Which boards have hardware RAID built in?

Software RAID comes in two flavours - BIOS level and OS level.

Hardware RAID comes on add-in cards (for the likes of LSI etc.) and requires Enterprise class hard discs.

[+Byt Subscriber¹]

I meant BIOS level sorry, my bad :) I tend to stay away from Software RAIDS as OS level.

When i said motherboard with built on RAID, i mean in the BIOS :blush:

[n_K]

Hardware RAID doesn't require enterprise class hard drives, it's better if you do but anyone can get their PC, buy a PERC5/i off ebay with an SAS -> SATA breakout cable and plug 4 or 8 SATA hard drives into it and it'll work fine.

[McKay]

Im getting a new laptop that has 2 x 500GB 7200RPM drives in RAID 0, what kind of speed increase will this yeild over a single 7200RPM Drive?

[Vice]

It's funny when RAID first came out the idea was to use commodity consumer hardware with it. RAID itself stands for Redundant Array of Inexpensive Disks. But as time has gone on using these inexpensive disks has become troublesome with manufactures including recovery routines on drives that stop the disk from communicating with its controller while attempting to access data in a damaged sector. Sometimes for over a minute. These recovery routines can make the RAID controllers (hardware ones specifically) believe that the disk has failed and drop it from your array.

If you're using RAID0 this would be the end of your data. If you are using RAID5 you now have to perform a rebuild which could take a while and if another error is found during this rebuild process and the same thing happens again with your remaining disks you will find your array destroyed. A similar situation can happen with RAID6 but you'd need three drives to fail in this manner so the risk is minimized but still a concern.

I've been using RAID for the past 8 or so years with different levels of success. Choose a good hardware controller or a good software solution like ZFS and combine that choice with a good set of drives that are known for their reliability and you'll have a good experience, use a cheap controller or just onboard motherboard RAID or a poor software solution like Windows software RAID5 and you will have a poor experience, slow, constraints on array expanding, mode mitigation, repairing or all of the above.

And something I always feel is important to say when these threads crop up is do not use RAID as a backup solution. RAID1, 5 and 6 may sound like they are designed for safeguarding your data and in some ways they are but they are not backup systems. They are time savers. And that is how you should treat them, if you have 10-12TB of data (as an example) it might take you a week to pull all data from backups and put it back where it belongs. RAID helps to mitigate that time by using some of your storage in your RAID array for redundancy against disk failure but if that fails (and it very easily could) you will be left to using your backups. Thus it's just a time saver and not something you can or should depend on in a crisis, it's just a convenient safety net, but it's a poor quality net and you could still fall through!

Im getting a new laptop that has 2 x 500GB 7200RPM drives in RAID 0, what kind of speed increase will this yeild over a single 7200RPM Drive?

In a laptop like that the access latency (time it takes to find and open a file etc) will increase up to 2x that of a single drive. In reality it will probably increase 15-25% - But throughput will increase up to 2x that of a single drive as well. So if a single drive has an access latency of 10ms and a throughput of 90MB/s it could end up around 12.5ms and 160-180MB/s. This is all hypothetical numbers since I don't actually know what drives you'll end up with but that is basically the sort of thing you can expect.

[+Byt Subscriber¹]

And something I always feel is important to say when these threads crop up is do not use RAID as a backup solution. RAID1, 5 and 6 may sound like they are designed for safeguarding your data and in some ways they are but they are not backup systems. They are time savers. And that is how you should treat them, if you have 10-12TB of data (as an example) it might take you a week to pull all data from backups and put it back where it belongs. RAID helps to mitigate that time by using some of your storage in your RAID array for redundancy against disk failure but if that fails (and it very easily could) you will be left to using your backups. Thus it's just a time saver and not something you can or should depend on in a crisis, it's just a convenient safety net, but it's a poor quality net and you could still fall through!

This will make you cringe....

I use to work for a company who would quote for a server for a client and insist that the RAID is their form of backup, they would actually tell the client that to check the backups they must check to see if all the hard drive led's are functioning on the drive bays and none are red.

Its safe to say.. i no longer work for that company lol

[Detection]

I ran RAID 0 for a long time and loved it, RAID 0 combines 2 drives (or I think you can add more) to give you their combined storage capacity and double their r/w speeds, but not only that, I found my system no longer locked up under heavy drive use any more

It brings with it the chance you could lose data if 1 drive fails though as half of every file is written to each drive, so if one drive dies, you are left with 1 drive holding only half of every file... not great but I never had a failure and never stored anything I couldn't get back on those drives anyway

Makes a huge difference to the performance of the system imo, I went from 2 x 160GB drives in RAID 0 to 1 x 1TB SATA II drive and felt like someone had taken half my specs out my machine

[+Fahim S. MVC]

Hardware RAID doesn't require enterprise class hard drives, it's better if you do but anyone can get their PC, buy a PERC5/i off ebay with an SAS -> SATA breakout cable and plug 4 or 8 SATA hard drives into it and it'll work fine.

Not sure what SATA/Not SATA has to do with enterprise classification as there are enterprise class SATA drives available. Of course what you are suggesting will work, but will be more prone to drive drop-out and cascading rebuild failures defeating the point somewhat. Read up on TLER or CCTL.

[n_K]

Not sure what SATA/Not SATA has to do with enterprise classification as there are enterprise class SATA drives available. Of course what you are suggesting will work, but will be more prone to drive drop-out and cascading rebuild failures defeating the point somewhat. Read up on TLER or CCTL.

TLER looks to be propietry to western digital hard drives, and the fact you can only configure it from DOS means it's pretty stupid. CCTL is the equilivent for samsung/hitachi but the major H/D players are WD, seagate and fujitsu, so really only 1 of 3 offers the ability to configure it.

Also, what I meant was SAS vs SATA has different communication and connectors.

There's plenty of people and places using high-end RAID cards with general hard drives, the poster here is obviously asking about a home setting, not a mission critical server for a bank with unlimited funds to spend on expensive hard drives to be utilized in RAID best :p

[Vice]

Actually you don't need to configure TLER timeouts from DOS. You can use SmartCTL from within Windows to set the value on Western Digital, Samsung and Seagate disks. And it works regardless of the RAID card you are using. Which obviously is important!

Also SATA and SAS use the same connector except SATA has a gap between the Power and SATA connector and SAS does not. The real difference is that SAS uses the SCSI communication standard and has a higher voltage for data integrity over longer distances. This is important obviously in server environments where you may have disks in one cabinet and the disk controller in another. You can safely use SATA disks over SAS SFF-8088 and 8087 cables but using SAS drives on SATA equipment will result in nothing happening because whilst most SAS controllers support SATA, the reverse is not true for SATA controllers.

You may wonder if these are the only differences then why do so many 14K and 15K hard drives only come in SAS versions? Well the reason for this is quite simple, the customers for these disks are using them in server environments so it makes sense to use the interface that is designed for the market you're targeting.

Personally in my own server I use a SAS RAID controller and an SAS Expander which takes Four SAS channels from my RAID controller and expands those in to 32 channels. I then run 30 Disks on that expander by using SFF-8087 to SATA breakout cables (4 SATA connectors per cable with one SFF-8087 on the other end which plugs in to the expander). The only part that is SATA in my configuration is the hard drives and everything else is considered SAS equipment.

[metro2012]

ive been using RAID5 for years and it has saved my ass countless times (not only that but i get free hard drives because of warrenty).

IMO, the biggest con for raid5 is the one im finding out now: expanding it. to expand it, you basically have to buy 4 disks at a time which is expensive (upgrading my raid5 to 3tb x 4 is going to cost me about 550 euros).

go for the largest raid5 you can; if not you might regret it later, like im doing now....

[Vice]

ive been using RAID5 for years and it has saved my ass countless times (not only that but i get free hard drives because of warrenty).

IMO, the biggest con for raid5 is the one im finding out now: expanding it. to expand it, you basically have to buy 4 disks at a time which is expensive (upgrading my raid5 to 3tb x 4 is going to cost me about 550 euros).

go for the largest raid5 you can; if not you might regret it later, like im doing now....

You should only need to purchase one extra drive to expand your RAID array. If your controller requires you to add four extra disks then there is something very wrong with that controller and I would recommend changing to something else. I've used HighPoint, LSI and Adaptec cards and all three of those have supported capacity expansion by adding a minimum of one new drive to the array. They also have all supported adding 2-3-4+ drives at once.

The only reason your card would ask for more disks is if you were changing raid modes from say RAID0 or RAID1 to RAID5 or from RAID5 to RAID6 while simultaneously expanding your array size at the same time.

[iuerg87yerg879e0rg9erugjer]

i am surprised no one has mentioned raid-z (i am running raid-z level 1 which is equivalent to raid-5 on my hp microserver with 12TB of space) or even zfs for that matter. I think is absolute brilliant and is extremely underrated.

here is a video if you have not herd of zfs before (skip to about 27mins for zfs)

[+LogicalApex MVC]

You should only need to purchase one extra drive to expand your RAID array. If your controller requires you to add four extra disks then there is something very wrong with that controller and I would recommend changing to something else. I've used HighPoint, LSI and Adaptec cards and all three of those have supported capacity expansion by adding a minimum of one new drive to the array. They also have all supported adding 2-3-4+ drives at once.

The only reason your card would ask for more disks is if you were changing raid modes from say RAID0 or RAID1 to RAID5 or from RAID5 to RAID6 while simultaneously expanding your array size at the same time.

I'm not sure I understand what you're saying here Vice. I have an LSI RAID controller and it will similarly not allow me to do an OCE (Online Capacity Expansion) beyond the size of the smallest drive in a RAID 5 array. So, if I had a 3x320GB RAID 5 array that I intended to move to a 3x1TB configuration I would first need to replace all three drives (in a one at a time then rebuild manner) before I could expand the array to use the new capacity.

There is no possible way to replace one of the 320GB drives with a 1TB drive and then expand the size of the array while the other two drives are still 320GB drives. If there was, where would the parity data go? The space just physically doesn't exist on the other two drives in the array.

[Vice]

I'm not sure I understand what you're saying here Vice. I have an LSI RAID controller and it will similarly not allow me to do an OCE (Online Capacity Expansion) beyond the size of the smallest drive in a RAID 5 array. So, if I had a 3x320GB RAID 5 array that I intended to move to a 3x1TB configuration I would first need to replace all three drives (in a one at a time then rebuild manner) before I could expand the array to use the new capacity.

There is no possible way to replace one of the 320GB drives with a 1TB drive and then expand the size of the array while the other two drives are still 320GB drives. If there was, where would the parity data go? The space just physically doesn't exist on the other two drives in the array.

In his post he never mentioned anything about changing his drives, only about expanding his array. So my post was regarding how you expand your array by adding more of the same sized disks.

I was not suggesting you can add different sized disks to an array to get more storage, I was saying you can use OCE to add one extra drive to an array and he made it sound like you have to always add 4 extra drives to an array to expand its size. He should have been more clear about actually wanting to change from four lower capacity disks to four higher capacity disks.

[Kryspy]

I am using SnapRAID myslef. Although not a true RAID solution; more of a Snapshot RAID it works quite well and doesn't require your drives to be formatted before using.

http://snapraid.sourceforge.net/compare.html

Kryspy

[metro2012]

You should only need to purchase one extra drive to expand your RAID array. If your controller requires you to add four extra disks then there is something very wrong with that controller and I would recommend changing to something else. I've used HighPoint, LSI and Adaptec cards and all three of those have supported capacity expansion by adding a minimum of one new drive to the array. They also have all supported adding 2-3-4+ drives at once.

The only reason your card would ask for more disks is if you were changing raid modes from say RAID0 or RAID1 to RAID5 or from RAID5 to RAID6 while simultaneously expanding your array size at the same time.

?

if im running a 120gb + 120gb + 120gb + 120gb RAID5 and I want to expand it to 1TB I have to replace ALL the disks because a RAID5 limits itself to the lowest capacity

120gb + 120gb + 120gb + 1TB RAID5 would be the same thing as 120gb + 120gb + 120gb + 120gb but that last disk would have a lot of unused space.

In his post he never mentioned anything about changing his drives, only about expanding his array. So my post was regarding how you expand your array by adding more of the same sized disks.

I was not suggesting you can add different sized disks to an array to get more storage, I was saying you can use OCE to add one extra drive to an array and he made it sound like you have to always add 4 extra drives to an array to expand its size. He should have been more clear about actually wanting to change from four lower capacity disks to four higher capacity disks.

i thought it was pretty clear

IMO, the biggest con for raid5 is the one im finding out now: expanding it. to expand it, you basically have to buy 4 disks at a time which is expensive (upgrading my raid5 to 3tb x 4 is going to cost me about 550 euros

[Vice]

?

if im running a 120gb + 120gb + 120gb + 120gb RAID5 and I want to expand it to 1TB I have to replace ALL the disks because a RAID5 limits itself to the lowest capacity

120gb + 120gb + 120gb + 1TB RAID5 would be the same thing as 120gb + 120gb + 120gb + 120gb but that last disk would have a lot of unused space.

i thought it was pretty clear

No it wasn't clear because you said expanding. Expanding an array is completely possible if you use the same sized disks or larger disks. It is known as Online Capacity Expansion or OCE and it is common in RAID language to say Expanding when adding an extra disk to a pre-existing array.

If you had 250GB + 250GB + 250GB + 250GB and then added a 1TB disk to that array you can expand that array. But that array won't be expanded by 1TB, only an extra 250GB with the remaining 750GB of that 1TB disk going unused.

What you should have said in your original post was that you wanted to change from lower capacity disks to higher capacity disks you then have to replace all your old disks in the array. Saying expanding the array makes it sound like you are unable to expand the array in any way at all.

In your post here:

IMO, the biggest con for raid5 is the one im finding out now: expanding it. to expand it, you basically have to buy 4 disks at a time which is expensive (upgrading my raid5 to 3tb x 4 is going to cost me about 550 euros

You never mentioned what your current RAID array was made up as. I believed from your post that your RAID5 array was already made up of 3TB disks and you merely wanted to add more disks to it and you were saying that you had to add another 4x3TB because your RAID controller wouldn't allow you to only add 1x3TB to your pre-existing array. You didn't make it clear that your array was using much smaller disks and you actually wanted to change disks.

[+Fahim S. MVC]

If you had 250GB + 250GB + 250GB + 250GB and then added a 1TB disk to that array you can expand that array. But that array won't be expanded by 1TB, only an extra 250GB with the remaining 750GB of that 1TB disk going unused.

To me this is one of the major downsides of RAID - I am seriously looking at the likes of ZFS or Storage Spaces as a result (I have 4x2.5 drives and 4x3.5 drives).

Does anyone know if there is a hardware solution that has the capacity utilisation effectiveness of ZFS or Storage Spaces?

[n_K]

It's an unavoidable downside because you don't have a spare 500GB or whatnot on the other drives for the RAID-5 redundancy data.

If you COULD use all the data on the 1TB disk and it failed, you wouldn't be able to put another 1TB disk in and rebuild the array, your data would be gone because it lacked redundancy data.

[Vice]

To me this is one of the major downsides of RAID - I am seriously looking at the likes of ZFS or Storage Spaces as a result (I have 4x2.5 drives and 4x3.5 drives).

Does anyone know if there is a hardware solution that has the capacity utilisation effectiveness of ZFS or Storage Spaces?

I don't know of any hardware raid cards which can support what you want. But to be honest ZFS is better anyway. I've read quite a few studies on the reliability of ZFS and in every one it has been found to be much more reliable than hardware RAID cards.

So it has a lot of benefits such as..

File Deduplication (Although this needs a lot of RAM and isn't recommended to use right now)

3 Disk Parity instead of dual like RAID6

Use different sized disks

Constantly checks the least accessed sectors of your disks for consistency so data corruption doesn't creep up on you

A lot more cost effective because you can use cheap interface cards instead of expensive raid cards

Doesn't care if disks have TLER or not because it will happily wait for a disk to complete its 1+ Minute recovery of bad sectors without dropping the disks from the array.

With all these pros the only reason not to use ZFS is that choosing it pretty much dictates your chosen operating system to FreeBSD, FreeNAS or Solaris. Apart from that it is the perfect storage system.