Is Raid5 still better than non-raid?

theinv

Limp Gawd
Joined
Apr 17, 2008
Messages
216
With today's high capacity drives, hard drive failure happens more frequently then it was 5-10 years ago. Is Raid5 still better than non-raid?


Consider these two scenarios:

case 1: with full external backup

1) 10 x 2TB RAID5

With this setup,

-- losing 1 drive will require a time consuming rebuild
-- losing 2 drives, all data is lost and I need to replace the 2 drives and then copy the data over from my backup.

2) 10 x 2TB non-raid

With this setup
-- Regarless of how many drive dies, I only need to replace the dead drive and then copy the lost data from my backup. It is fast and efficient

​



case 2: no external backup

1) 10 x 2TB RAID5

With this setup,

-- losing 1 drive will require a time consuming rebuild
-- losing 2 drives, all data (18TB) is lost

2) 10 x 2TB non-raid

-- With this setup, I only lose data on the hard drive that is dead.

​


In either case, the non-raid appears to be a better solution. I understand you can use RAID6, but RAID6 will have same problem. What if 3 drives die at the same time?....

What do you think?
 
RAID 5 rebuilds don't take THAT long if you have dedicated hardware. Last time I did it it took 18 hours on a 4-drive, 4.5TB array with a PERC 5/i. Considering you can have them start automatically and run all night, it's not really "time-consuming". I prefer that flexibility of being able to lose one drive but no data.

Also, if you can do a 1-to-1 backup of your entire giant array, you're not in the majority :rolleyes: I know it's a sin, but my RAID 5 is my backup plan, because if I had the money for another ~4 1.5 TB drives I'd have a 9TB RAID 6 array already ;)

EDIT: And yea I forgot about the read speed bonus of RAID 5 as StarTrek4U said.
 
The likelihood of 2 drives failing at the same time is pretty low unless you have some sort of crazy power issue or something.

You need to be more specific when you say "non-raid" Are you saying you're going to have those drives just independent of each other? That's horribly inefficient since you lose the excess space and each drive and trying to find/remember where you put your data would be a huge PITA.

How are you backing up all this data? How easy is it to get back from a backup? Perhaps the most important question- is this for home or business? If it's for a business RAID wins, end of story since they redundancy/uptime/managment is a must have in the working world.

Also, in RAID setup you're striping data which gives you better speeds across a large number of disks than if you leave them independent of each other. You don't specify the controller you're using (which is also important) but most decent ones will do rebuilds on the fly so if/when you lose a disk you can still access your data while the array rebuilds.
 
Murphey's law says that the drive you loose will be the one you least want to loose. At least with a RAID solution there is a chance of recovery, and with hot spares the time to recovery can be mitigated.

Right now, I am trying to get 4 x 2TB drives to stick in a RAID 10 array to serve as storage.
 
Last time I did it it took 18 hours on a 4-drive, 4.5TB array with a PERC 5/i

That seems horribly slow. It takes me about 8 hours for the same on linux software raid.
 
That seems horribly slow. It takes me about 8 hours for the same on linux software raid.

It was, not sure why... I guess because they're WD greens. When I had my Samsung F1 750's a 5-disk rebuild took me just under 9 hours. That said, though, I've only ever done one rebuild with each (adding additional drives), so it could have been a fluke with other factors involved. My point about it not really being THAT long still stands though ;)
 
The whole point is that RAID is NOT a backup. Say this 10 times fast. The whole point is that, with a proper backup strategy in place, you will have high availability to your data. If you have RAID5 and 1 drive fails (or up to 2 with RAID6) you will not have to go without at all. Your data is still up and running for your users, saving them downtime while you restore the data. You also don't lose the data between the time of your last backup and the drive failure. It is just one part of many for data availability and sustainability.
 
Rebuilding a ZFS takes much less time; with an empty filesystem its instant, so it depends on how much data you have stored.

If you include ZFS/RAID-Z in your comparison, it would also be much safer than traditional RAID5.
 
That seems horribly slow. It takes me about 8 hours for the same on linux software raid.

From the numbers I've seen other people posting for rebuild times on hardware controllers, they are very slow compared to Linux softraid at performing rebuilds and especially OCE/level migration.

A 10 disk RAID5 is borderline nuts, especially with 2TB disks. Use RAID6 or two 5 disk RAID5s.
 
With today's high capacity drives, hard drive failure happens more frequently then it was 5-10 years ago.

LOL. Have you conducted a study? My experience has been just the opposite. The drives of 5-10 years ago were ridiculously unreliable, and I had way more "click of death" situations happen out of the blue than on drives owned within the last 5 years.

It's also news to me that high capacity (platter density) is linked to do more frequent failure rates. Do tell.

I understand you can use RAID6, but RAID6 will have same problem. What if 3 drives die at the same time?....

What if you get hit by lightning three times on the same day? It could happen, right.

The only way 3 drives die at the same time is if they're being abused, like being run in an environment with excessive temperature / inadequate ventilation, or you're using a crap power supply, etc. Drives running within normal operating conditions don't die 3 at a time, let alone 2 at a time.
 
Last edited:
Why not RAID 6?

A 8x 2tb RAID 6 on an Adaptec 5805 takes maybe 12 hours, at tops to rebuild plus you have 2 drive redundancy.
 
The only way 3 drives die at the same time is if they're being abused, like being run in an environment with excessive temperature / inadequate ventilation, or you're using a crap power supply, etc. Drives running within normal operating conditions don't die 3 at a time, let alone 2 at a time.

odd-
I have to disagree with you here. I have had an HP Proliant DL360 power supply partially "just go bad" and fry 4 drives out of an array (server otherwise stayed up). Sometimes shit just happens, which is why you need a good backup plan as part of any data storage.
-Mike
 
Agreed and I've had same happen in some old ML580's and even 8 drives in a home raid get hosed by a PSU all at once, but that was my point - that 3 drives wouldn't fail on their own if all other variables like power and temps are nominal.

Also, no one's arguing you need a good backup plan. Raid versus non-raid was the original question raised by the OP, and as you know backup is a completely separate issue. RAID is about performance and reliability. Not backup.
 
Back
Top