On 04/18/12 03:28 AM, Roger (K8RI) wrote:
> On 4/17/2012 6:35 PM, Dr. David Kirkby wrote:
>> On 04/14/12 11:34 PM, rick darwicki wrote:
>>> Anyone had any problems with the new solid state computer drives? Blowing
>>> them away with RF or RFI ???
>>>
>>>
>>> Rick, N6PE
>> From what I gather, they are pretty unreliable - it does not need any
>> RF around
>> them! From what I can gather, they are far less reliable than hard disk
>> drives.
>
> I have 3 machines with them and will probably have a 4th soon. The 4th
> fails to boot on over 75% of cold boots and comes up with the error
> message to insert the system disk and run repair. If I just reach up
> and hit "reset" it reboots quickly. Shutdown and restart also works. I
> don't remember it failing yet. I have a feeling there is a problem with
> the mechanical C drive spinning up. As I need to swap out the drive to
> prove it anyway I might as well do it with a good SSD and SSDs are
> available in a wide range of quality and speed. They are not a place
> to buy cheap. You need to find one with good specs and good
> reliability. If for nothing else go to NewEgg and read the user
> write-ups on them. Some are good and some are not so good. Like on here
> there are those with experience and those with out including those
> writing reports who've never owned one.
I take that as meaning me, since I've never owned one.
An individual writing about their own reliability experiences is not very
useful, unless that individual has used a very large number of them, as some do
in data centres. Statistically it is pointless. But I've certainly seen
evidence
from users who have used a lot of them, that the failure rates are unacceptably
high.
It's only sensible to consider failure rates per GB of stored data, NOT in
absolute numbers of disks, yet that is what you are doing.
If a typical SSD disk is 128 GB, and a typical mechanical drive is 1 TB, then
clearly to achieve the same data storage, you are going to need 8 times as many
SSD disks as one mechanical disk. One might reasonably expect storage devices
to
work for 5 years. (Some might say 3 years, some might say 10, but lets settle
on
5 for an example).
If the probability of failure of one storage device inside 5 years is 'p', then
the probability of one disk surviving 5 years is (1-p). But the probability of
N
disks surviving 5 years is much lower at (1-p)^N.
If 1% of 128 GB disks fail within 5 years, 99% survive 5 years. But the
probability of you keeping the same 1 TB of disk storage safe for 5 years is
just 0.99^8 = 0.922 = 92.2%.
To put that in simple terms, for SSDs to be as reliable for data storage as
mechanical drives, their failure rates need to be about 1/8th that of
mechanical
drives. That is simply not the case.
The PC I'm using to type this email is a bit untypical, as it's a Sun
workstation running Solaris, although its an x86 Xeon processor, so could run
Windows if I wanted.
I have a pair of 2 TB enterprise grade SATA disks for data storage, mirrored.
So
if either one fails, my data remains intact. It would be extremely expensive to
buy SSD disks to store 2 TB, while have redundancy. I'd need to buy 32 x 128 GB
SSD disks.
Dave
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|