Re: Colorsync-users Digest, Vol 10, Issue 65
Re: Colorsync-users Digest, Vol 10, Issue 65
- Subject: Re: Colorsync-users Digest, Vol 10, Issue 65
- From: Morgan Gordon <email@hidden>
- Date: Tue, 30 Apr 2013 15:29:00 -0700
Graeme, I haven't heard any horror stories, but even if I had they'd merely be anecdotes.
> And it's getting worse as the manufacturers try and cram more data in smaller and smaller cells.
> Some of the earlier Flash devices took 10000 write cycles or so to wear out, but this is dropping -
> there's talk of making SSD's out of flash that wears out in as little as a few hundred writes !
My understanding is that MLC NAND has a much shorter lifecycle than SLC NAND. At the beginning of SSD market penetration, all SSDs utilized SLC NAND cells which could write to one block at a time versus MLC NAND cells which required writing to five blocks at a time even if you only needed to record one blocks worth of data.
I do agree that hard drives give some warning. Just the other day, one of my seagate backups was running at 85 degrees C... I returned it.
Overprovisioning does do the trick and I don't think of it so much as a hack to make them stable as a necessary consequence of writing to a medium inherently less stable. The overprovisioning is monitored in SMART data and that, I think, is an adequate means of safetly establishing their remaining lifespan.
On most MLC, cheaper SSDs, the overprovisioning is just a few percent of the total memory on the drive. With better companies such as Samsung, OCZ, and OWC, you can expect the drive to have between 5 and 7 percent over provisioning, potentially giving you an MTBF of over 3 million hours. And enterprise caliber SSDs that will really destroy your budget from companies like HP, Seagate, and OWC will use upwards of 12% of the total memory for over provisioning. The worn sectors are presented in the SMART data for the drive, so you can tell when it's reaching end of life. That said, each drive relies on a microprocessor to handle all writes and reads from the drive, abstracting block level access from the computer. If you were to have a failure in the processor or lose power before a write operation had finished, it could conceivably result in an immediate loss of the device. MTBFs for enterprise level SSDs exceed those of nearline SAS HDDs. Most of us use neither though, unless we're working with a large company and leveraging servers. I've had a lot of experience with HDDs and SSDs but am certainly not an expert on the topic. Of the six SSDs I've purchased in my life since they first became readily available in the 2007-2008 (?) range, I've never had a failure even when using them as personal servers that were constantly being written to.
Whereas, the longest lived HDDs I currently own are the ones in my Promise Pegasus, I've had such bad luck with HDDs and especially the 2.5 inch laptop (not enterprise) variety that even if presented with a study that demonstrated with 6 sigma certainty that HDDs were longer lived and more reliable than SSDs, I would honestly probably rationalize the results away as errors in testing methodology. To this day, the price per megabyte ratio is still inferior for SSDs. But right after the floods in the far east where most HDD manufacturing occurs, SSDs became much more competitive.
I don't just do still photography. I do a lot of video work. Purely as a function of time, I can significantly increase my productivity during video ingest achieving sustained transfer rates on my two year old laptop in excess of 250 MB / S. They're silent, typically (though not always) draw less power than comparable HDDs and arent subject to problems as a result of head to platter contact during movement which is a real concern with a laptop you're moving while using despite the laptops sudden motion sensor that parks the hard drives in anticipation of a shock that would damage the HDD.
Finally, running a web browser, lightroom, photoshop, the OneOne suite, DxO, Bridge, and various plugins fragments my memory and for whatever reason results in progressively less stable systems despite the solid underpinnings of darwin. When I can restart my computer and achieve a relatively fresh state or, in a worst case scenario, start in single user mode, run fsck, repair permissions, and then restart again, all in less then 20 seconds, I feel much less stressed than I do with an HDD based system where the process of restarting and shutting down can sometimes takes five or more minutes. Things become more responsive and as such, I can rule out the causes of a slowdown or error much more quickly.
This is a colorsync listserve though and I readily admit that my SSDs have no influence whatsoever on how well my spectro calibrates the limited gamut on my internal Mac Display. :P
Morgan
P.S. Graeme, until recently, SSDs did have a horrible reputation for not being a secure storage medium. The technology has advanced quickly both in garbage collection (the continuous process that manages the distribution of data on the NAND chips for longevity and to maintain speed) and the quality of the NAND memory integrated into most of these devices. For my mother's computer, I was able to outfit her with a SanDisk 480 GB SSD for $300 dollars. I put it in her 2007 mac mini and even though the computer is limited to 2 gigabytes of ram and she doesn't know how to close tabs in safari, the increased responsiveness of the system (I attribute it to the OS using the SSD for paging files) is so much greater than it was prior to the upgrade that she decided against buying a new iMac and instead spent the money on a really, really old and used RV... now that I think about it, maybe I did her a disservice :P
> Message: 4
> Date: Tue, 30 Apr 2013 08:50:37 +1000
> From: Graeme Gill <email@hidden>
> To: ColorSync <email@hidden>
> Subject: Re: Colorsync-users Digest, Vol 10, Issue 64
>
> Morgan Gordon wrote:
>
>> images I choose to retain that arent being worked on are removed from the SSD and spend their
>> days on the Pegasus. Then I have a seagate backup plus thunderbolt adapter and three 4 TB drives
>> that use with time machine to backup everything on the SSDs on the laptop and iMac and the
>> photos I keep on the Pegasus.
>
> Hmm. SSD has a bit of a reputation for being extremely fast, and quitting without any warning.
> And it's getting worse as the manufacturers try and cram more data in smaller and smaller cells.
> Some of the earlier Flash devices took 10000 write cycles or so to wear out, but this is dropping -
> there's talk of making SSD's out of flash that wears out in as little as a few hundred writes !
>
> Yes, they're pulling redundancy tricks (over provisioning and wear levelling) to cover this up,
> but I have a nasty suspicion that there will be an assumption that "you'll buy a new machine/SSD
> every two years anyway", so that it "doesn't matter" if new SSD's fall over very quickly.
>
> In contrast, hard drives typically give some warning when they are on the way out.
>
> Graeme Gill.
>
>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden