I doubt it's TRIM or anything like that. There were some problems years ago with TRIM implementations, but recently it seems a bit better.
I'm guessing the AtomRAID basically wraps a case around a couple 2TB M.2 NVMe drives/sticks, and then uses RAID 0 to combine the two higher-speed buffers in each NVMe for speed (duh) and to quickly transfer bigger file sizes (I think).
The performance degradation could happen when you're reading and writing more data than the drive caches can hold (and as they get so full, I think). As happens when editing. Check out the charts on this page (and click through their links for deeper info). The article might be about a different memory technology than Glyph is using, but conceptually, I think the idea's the same. Just scroll down to the graph about half-way down... Notice the huge performance drop: https://www.howtogeek.com/428869/ssds-are-getting-denser-and-slower-thanks-to-qlc-flash/
I've been backing up cameras in the field to SSDs since last summer; not so much spinning disks these days, mainly for the speed issues Wilson alludes to. I'm using little 2TB SanDisk externals. They've introduced faster models, but mine are fast enough for the cards we're using. https://shop.westerndigital.com/products/portable-drives/sandisk-extreme-usb-3-1-ssd#SDSSDE60-2T00-G25
I hope your producers can get their files off their drives. If they do, perhaps they'll be ready to improve their working protocol.
Good luck John.
> On Jan 17, 2020, at 8:21 AM, John Moore <bigfish@pacbell.net> wrote:
>
> No argument other than with spinning drives it's worked for the company. When you say it's uncommon to be deleting and recovering space I'm not sure I'm understanding. My limited google lead to the way an SSD has to rewrite a whole block, which can be up to 128 pages (4KB/page) so changing just a page means caching a whole block and rewriting it back with the modification of the new info. I read some approaches will keep empty blocks so they don't have to delete the existing block but just write to an empty one and flag the existing block for deletion. I'm not clear as to the methods of how these processes are administered and if it's up to the OS or does the SSD itself keep track of this stuff.
>
> I'm grabbing at straws here but the way the drive problem was described to me it was progressively worse which sounded like the types of performance degradation of SSDs I'd read about. Still degradation doesn't mean failure so I'm really at a loss as to what is really happening under the hood.
>
>
> On Thu, Jan 16, 2020 at 11:31 PM, Pat Horridge wrote:
> Who edits onto the camera sources drive (beyond fast turnaround news)
> Madness. And who writes to a drive with just 5% free space. Again madness.
> However as to multiple writes being the issue I suspect not as it's uncommon to be deleting and recovering space. Who does that nowadays.
> Sounds like a faulty batch.
> But so training on good working practices would be helpful.
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#133882): https://groups.io/g/Avid-L2/message/133882
Mute This Topic: https://groups.io/mt/69738311/2971402
Group Owner: Avid-L2+owner@groups.io
Unsubscribe: https://groups.io/g/Avid-L2/leave/5759084/1548653033/xyzzy [administrator242.death@blogger.com]
-=-=-=-=-=-=-=-=-=-=-=-
No comments:
Post a Comment