[dba-Tech] SSD Diagnoser

John Colby jwcolby at gmail.com
Mon Sep 12 16:30:17 CDT 2016


 >So, all things being equal, SSD drives should last as long as any 
traditional spinning-rust drive?

Yes and no.  In some applications such as high transaction web sites or 
databases they will definitely not last as long, simply because these 
activities write a lot and very often.  But they will provide 10X or 
better speed increases.  I have read about data centers changing out 
their SSDs every 90-180 days because they had so much data written to 
them.  But hey, they still use them because of that blazing speed.

For the desktop, my answer would be yes, as long.

For laptops, my answer would be "probably longer than spinning rust".  
The reason being that moving a disk drive, particularly twisting it 
around the spinning axis is hard on drives.  Think Gyroscope.  Dropping 
disk drives is hard on them.  Neither of these things will even be 
noticed by an SSD since it is after all "solid state".

In fact for my laptops, the spinning drive lasted just long enough to 
order an SSD and install it!  So the SSD already has outlasted that 
spinning disk.

:-)

On 9/10/2016 11:42 PM, Jim Lawrence wrote:
> Thanks for your experience.
>
> So, all things being equal, SSD drives should last as long as any traditional spinning-rust drive?
>
> Jim
>
> ----- Original Message -----
> From: "John Colby" <jwcolby at gmail.com>
> To: "Discussion of Hardware and Software issues" <dba-tech at databaseadvisors.com>
> Sent: Thursday, September 8, 2016 8:20:29 PM
> Subject: Re: [dba-Tech] SSD Diagnoser
>
> Jim,
>
> As I repeat I am not an "expert", but your are in a way correct. SSDs
> have a /dedicated/ section of memory (the disk) used for repairing the
> disk, and yes, the more expensive the SSD the larger this section.  Thus
> if the disk is completely full, then the more expensive disk can "self
> repair" longer because of this larger dedicated area.  However, if the
> disk is not completely full, then the controller can simply move data
> out of sectors which are dying into other sectors which are fine and map
> the bad sectors, in the same manner that a rotating disk will do.
>
> As I used in my example, suppose I have an MDB file which gets written
> to a lot, and a huge PDF file which is written once and just sits
> there.  (As I understand it) the controller will figure out that the MDB
> file is "wearing out" the sectors it sits in, and can simply "swap" the
> MDB file with PDF file.  The memory cells that the PDF has been sitting
> in are absolutely brand new, written once.
>
> You can see how this concept can "level" the wear of the disk, even
> without dipping into the dedicated areas used just for such "repair".
>
> SSDs are actually rated in total Terabytes written, and / or Gigabytes /
> day written.  As I visualize it, the disk is essentially "used up"
> (eventually) as the wear leveling algorithms moves high wear files into
> low wear areas.  Modern SSDs are pretty much not going to wear out
> unless you are using them for high transaction databases or something
> like that.  IOW on the desktop they will probably never wear out.
>
> On 9/8/2016 10:48 PM, Jim Lawrence wrote:
>> I believe you are correct...sort of self-healing drives. But as I understand it there is a limit to how much a drive can self-repair...the more expensive the SSD the greater the limit.
>>
>> Jim
>>

-- 
John W. Colby



More information about the dba-Tech mailing list