Hi,

I have RAID on my computer and I have heard thay if one HDD fails, they both fail.

Should I get some backup (NAS or an external HDD) to be sure or is this type of set up quite reliable?

I have Hitachi HDD's (came with the computer)

Ben

Do you know how your RAID is configured as there are different levels? RAID 0, RAID 1 and RAID 5 being the common ones. I'm assuming yours is configured in RAID 0... Do you see both drives(assuming you only have two) as one drive?

With RAID 0, your data is striped across both drives so if one drive fails, all of your data will be lost. The other drive will not physically fail as well, but the data on it will be useless. People usually have this setup for it's performance benefits.

There's an explanation here: RAID - Wikipedia, the free encyclopedia

External storage has become quite cheap these days so I think it's well worth getting one to backup all your precious data.

It's RAID 0

Thanks for the help, I think I'll get an external storage just in case it ever goes t**s up.

Ben

It's RAID 0

Thanks for the help' date=' I think I'll get an external storage just in case it ever goes t**s up.

Ben[/quote']

I would get some sort of backup in place asap for those Deathstars... :eek:

As has been said , striped hard drives will almost double the read/write speed but also increases the chance of data loss.

I have this setup myself , and an external drive that I use as a backup

Yep, deffo going to get some backup storage now!!

Ben

Hi,

I have RAID on my computer and I have heard thay if one HDD fails, they both fail.

Should I get some backup (NAS or an external HDD) to be sure or is this type of set up quite reliable?

I have Hitachi HDD's (came with the computer)

Ben

If one fails they both fail is going to be RAID 0.

I really don't see the point in this set up for a home user as it is only faster than RAID1 in a limited number of specific write operations under certain conditions not usually encountered at home.

For a home user, I'd probably run 2 disks in RAID 1 and have a backup strategy for your important data. Say weekly or monthly onto DVD.

Be aware that you have to look after DVDs and they are not a long term storage solution.

I would get some sort of backup in place asap for those Deathstars... :eek:

The Hitachi drives don't deserve that name any more IMHO.

Obviously this is totally my opinion and has no connection with my job.

As has been said , striped hard drives will almost double the read/write speed but also increases the chance of data loss.

FWIW I have a lot of knowledge in this field and have the iometer and other data to prove otherwise.

The effect of Striping is a long way from doubling the write speeds (this depends largely on stripe size, access pattern and queue depth anyway). The read advantage is also not double that of a single disk.

The risk of failure however is potentially greater than double as you now have two drives with MTBF's to deal with. It's a common misconception that MTBF suggests the average life of a drive, but this isn't true. The link below helps to explain it better.

The Bathtub Curve and Product Failure Behavior (Part 2 of 2)

For that reasoning I'd totally agree with taking a good backup.

I run Raid 0 with my Laptop, but I back all my data up on to a terabyte drive. Admittedly all my settings and stuff like that would be lost, but all my Docs would be safe.

You have 2 drives in your laptop?

Also assuming you're using it for video streaming, that is one case where it can benefit

The Hitachi drives don't deserve that name any more IMHO.

Obviously this is totally my opinion and has no connection with my job.

Just basing that comment on my recent experience of them, which hasn't been great to say the least. Granted we're not talking hundreds of drives but of the few I've had the majority have failed.

Infact I have 2 Hitachi Deathstars sat on my desk atm that are less than 12 months old waiting to be RMA'd. :(

You have 2 drives in your laptop?

Also assuming you're using it for video streaming, that is one case where it can benefit

Yes, I have 2 x 160GB HDD's.

I do use it for streaming too

Ben

Hell..... I think I'll get some extra storage this weekend....the term 'Deathstar' is a bit worrying :eek:

Ben

Hell..... I think I'll get some extra storage this weekend....the term 'Deathstar' is a bit worrying :eek:

Ben

Sadly from my recent experience they've lived up to this name. :(

...It's a common misconception that MTBF suggests the average life of a drive, but this isn't true...

Sorry Mark, but since you've been quite vocal about the quality of advice posted I can't help but pick you up on this. If you have a population of lamps with a quoted mean time between failures of 1 year, then you can expect that the mean period of time before each lamp needs replacing will be one year. Some will last more; some less, and that's where reliability theory comes in.

While reliability theory states that the statistical service life of each of the lamps will be follow the Weibull distribution, all this means is that a proportion of lamps will suffer early-life failures (due to manufacturing defects etc.), then there is a period when the non-faulty lamps operate with an underlying reliability, and finally the lamps start to break down due to old age and thus the failure rate increases again. This comes into its own, therefore, when you need to specify proof-test intervals in systems suffering from unrevealed failure modes.

The point of a Redundant Array of Inexpensive Drives is that any one drive can fail, and you don't lose data. The broken, 'inexpensive' drive is taken out and replaced, and the system continues in the background without needing to be taken out of service. The debate whether RAID 0 can really be called 'RAID' isn't a topic for this thread, I think.

So back to lamps, if you took the mean service life of all of the lamps, you would have the MTBF. And if you picked a single lamp from the population, its service life would most likely correspond to the MTBF. So apart from reliability theory highlighting the importance of thorough commissioning, I don't think saying that MTBF and predicted average service life are completely unrelated is true. Fair enough, you can't choose a drive on the assumption that it will last at least as long as the MTBF, but that's a fundamental misunderstanding of statistics, rather than confusion between MTBF and predicted average service life...

Sorry Mark, but since you've been quite vocal about the quality of advice posted I can't help but pick you up on this. If you have a population of lamps with a quoted mean time between failures of 1 year, then you can expect that the mean period of time before each lamp needs replacing will be one year. Some will last more; some less, and that's where reliability theory comes in.

While reliability theory states that the statistical service life of each of the lamps will be follow the Weibull distribution, all this means is that a proportion of lamps will suffer early-life failures (due to manufacturing defects etc.), then there is a period when the non-faulty lamps operate with an underlying reliability, and finally the lamps start to break down due to old age and thus the failure rate increases again. This comes into its own, therefore, when you need to specify proof-test intervals in systems suffering from unrevealed failure modes.

The point of a Redundant Array of Inexpensive Drives is that any one drive can fail, and you don't lose data. The broken, 'inexpensive' drive is taken out and replaced, and the system continues in the background without needing to be taken out of service. The debate whether RAID 0 can really be called 'RAID' isn't a topic for this thread, I think.

So back to lamps, if you took the mean service life of all of the lamps, you would have the MTBF. And if you picked a single lamp from the population, its service life would most likely correspond to the MTBF. So apart from reliability theory highlighting the importance of thorough commissioning, I don't think saying that MTBF and predicted average service life are completely unrelated is true. Fair enough, you can't choose a drive on the assumption that it will last at least as long as the MTBF, but that's a fundamental misunderstanding of statistics, rather than confusion between MTBF and predicted average service life...

Really, so the mean time between failures for a hard disk is even close to the amount of time you can expect before a drive fails? Did you read the link to the information on MTBF I gave? The MTBF is very loosely connected to the expected life of the drive!

If a device has an MTBF of 1 year that doesn't mean it will last a year, far from it.

Most drives will have worn out well before the MTBF so you can not expect them to last that long.

I work in the storage field and have access to this data for quantities of millions of drives tested. I believe a very good amount of the storage arrays in the world are ours.

I'll drop you a PM and you'll probably realise why I'm quite so sure of this.

Really, so the mean time between failures for a hard disk is the amount of time you can expect before a drive fails hey? Did you read the link to the information on MTBF I gave.

If a device has an MTBF of 1 year that doesn't mean it will last a year, far from it.

Most drives will have worn out well before the MTBF so you can not expect them to last that long.

I work in the storage field and have access to this data for quantities of millions of drives tested.

I'll drop you a PM and you'll probably realise why I'm quite so sure of this.

OK! I'll read it and provide you with my response...

..

You have 2 drives in your laptop?

Also assuming you're using it for video streaming, that is one case where it can benefit

I do, 2 x 200, but I don’t do a huge amount of streaming. It just came that way and as I use externals storage fro most stuff I thought I would leave it alone. Mind you the motherboard blew up after a month when I got it last year…….it’s a Sony! My little Tosh seemed much more reliable if not as tank like in its build quality.

I was talking about the fairly heavy duty streaming, which i assume a sound engineer does but it's still quite rare to see two drives in a lappy these days.

FWIW on the MTBF line,

I can make a drive that i can 100% guarantee will die within a week and yet it could still have a huge MTBF ( say 1million hours).

The MTBF can not be used to determine reliability and it does not mean that 50% of drives will last that amount of time.

I was talking about the fairly heavy duty streaming, which i assume a sound engineer does but it's still quite rare to see two drives in a lappy these days.

.

On occasion I stream work stuff, but these days I seem to spend most of my time standing in muddy fields (it’s a glamorous and somewhat soggy life)

If the MTBF was true.. it would mean the new WD Velociraptor would last around 50 years....... ummm i doubt it

I probably ought to point out that Mark's and my little exchange seems to have boiled down to different methods of using failure rate data and how this is translated into reliability figures. Sorry for wasting all your time.

Just carry on - I've already got me coat...

Also with HDD, I reckon the life expetancy varies dramatically depending on the usage. As Mark said, it's not hard to put a certain HDD through a certain set of test and be pretty 100% sure you'll knacker it in a week.

The standard deviation of MTBF is rather large, and dependant on the aforementioned usage.