|
|
This topic comprises 3 pages: 1 2 3
|
|
Author
|
Topic: SSD drives in Doremi DCP2000
|
|
|
|
|
|
|
Leo Enticknap
Film God
Posts: 7474
From: Loma Linda, CA
Registered: Jul 2000
|
posted 02-13-2016 08:59 PM
One thing that would make me reluctant to try SSDs in a DCP server is that I'm guessing that the RAID controller (be that software or hardware) will be constantly shuffling data around to make optimum use of the space and minimize seek time, etc. The main determinant of a "spinning rust" HDD's lifetime is hours and spin up/spin down cycles. For an SSD, it's the number of times each sector is written. The more often you write to it, the more quickly you'll wear it out. The desktop operating systems (e.g. Windows and Mac OS) know this, and will not automatically defrag a SSD in the way that they will a traditional hard drive.
But RAID controllers are probably not designed with SSDs in mind: the point of a RAID is storage volume and data security, not speed of access and ruggedness, which are the two main reasons for using a SSD. So even if a DCP server's RAID controller would work with SSDs, I'd speculate that the risk is there that it would burn them out very quickly.
If I were operating a mobile DCP server in a portable rig, I think I'd be inclined to carry one or two spare hard drives around with me in the accessories case, and swap one in and rebuild the RAID in the event that one did not survive transit.
| IP: Logged
|
|
|
|
|
|
|
|
|
|
|
|
Marcel Birgelen
Film God
Posts: 3357
From: Maastricht, Limburg, Netherlands
Registered: Feb 2012
|
posted 02-14-2016 03:59 PM
quote: Leo Enticknap
One thing that would make me reluctant to try SSDs in a DCP server is that I'm guessing that the RAID controller (be that software or hardware) will be constantly shuffling data around to make optimum use of the space and minimize seek time, etc. The main determinant of a "spinning rust" HDD's lifetime is hours and spin up/spin down cycles. For an SSD, it's the number of times each sector is written. The more often you write to it, the more quickly you'll wear it out. The desktop operating systems (e.g. Windows and Mac OS) know this, and will not automatically defrag a SSD in the way that they will a traditional hard drive.
Well, most RAID systems do not constantly shuffle data around actually. MOST RAID systems and especially that of the average DCP server are quite dumb. If there is any shuffling, it usually takes place at the filesystem level. There are lots of (semi-)intelligent storage systems out there that do some preventive data shuffling, but they are not the ones you find in your DCI server.
Also, the kind of data you need for DCI is quite boring actually. Huge files, which usually get read and written rather sequentially. So unless you constantly ingest and play the data for like 5 screens simultaneously on the same machine, the level of fragmentation on your average DCI system is really a non-issue.
In case of SSDs, there actually is some black magic going on INSIDE the SSD itself. The controller on your SSD may remap your data as it sees fit. For example, if you keep on overwriting the same data, your SSD may start to write this data on some other physical location on the SSD itself, to keep the wear more balanced. Your RAID controller or your OS isn't even aware of this happening.
Shock resistance for hard disks actually should be a non-issue too, at least if you're not operating them while transporting them. The whole concept of DCI got started by shipping those disks around. Once your disks are in a parked state, forces that would damage the disk would also damage other parts of your equipment.
The biggest pitfall here is if you get those disks work reliably with the Doremi system. And although this is essentially a vanilla Linux software RAID solution, I'm not sure their layer around it will support it. So, you might end up with quite some hackery to get it rolling.
Another pitfall to account for is TRIM support. TRIM support is essential to avoid your SSD drives grinding down to a halt.
Although SSD drives are usually blazingly fast, one thing they're VERY slow at is erasing data. Unlike the rotating rust counterparts, SSDs cannot just overwrite data, they have to erase it first and that's a painstakingly slow process. Also, keep in mind that most delete actions aren't actual deletes, it's a mere "unlink" from the inode table. The problem with this process is that alhough your OS thinks it's "gone", the data is actually still there on your SSD.
So, once you've filled your disks to the brim at least once, you often hit this "nice" wall of total performance degradation.
That's where TRIM comes in. TRIM needs support from both your OS, your RAID and/or controller layer and your SSD. TRIM uses unspent I/O and system time to do some preventive housekeeping. Your OS tells your controller and therefore your disks what data is essentially gone. And in times of low system activity, the data gets prematurely erased by the SSDs, so once there is new data, it can be written without having to erase previously deleted data first.
Once you keep ingesting, you will hit this virtual performance rather soon, even if you keep deleting content. This "performance barrier" should not hurt playback speeds, but it will dramatically impair ingest performance and MIGHT even negatively impact system performance. So a correct working TRIM implementation is not just a nice to have for such a setup.
| IP: Logged
|
|
|
|
Leo Enticknap
Film God
Posts: 7474
From: Loma Linda, CA
Registered: Jul 2000
|
posted 02-14-2016 08:42 PM
quote: Marilu Magri
Thanks all for the input and ideas. As an alternative to SSD's, would a good alternative be to remove the raid drives and transport them in a padded container and then re-fit them at the show?
We have two festivals a year that are sponsored by Barco, and so we have to have a temporary loaner Barco projector shipped in for them. Boston Light and Sound takes care of the installation, operation and de-rig (hired by the fests), and they bring in a Doremi server to use with the Barco, to avoid having to reconfigure our DSS200 and provide a cat745 (and also because they can pre-ingest and test the content and KDMs, and arrive with everything ready to play).
Removing the hard drives for shipment is precisely what they do - in fact, they pack them in the same padded cardboard boxes that the drives ship from the Doremi factory in. I'm guessing that the hard drive failure rate they have from doing this is low enough for it to be a non-issue for them.
Even taking on board Marcel's belief that RAID controllers are capable of being kind to SSDs, my instinct on this is that the sheer cost of SSD storage versus spinning rust is not worth it for the ruggedness gain. And as he points out, probably around 99% of DCPs are shipped in traditional hard drives in padded boxes. In almost two years since I re-entered the cinema biz working in a DCP environment, I can only recall two DCP drives DOA (due to "click of death" drive failure) in all of that time, and both of those were 2.5" consumer drives still in their consumer retail packaging, not CRU drives in properly designed shipping containers.
| IP: Logged
|
|
|
|
|
|
|
|
|
|
All times are Central (GMT -6:00)
|
This topic comprises 3 pages: 1 2 3
|
Powered by Infopop Corporation
UBB.classicTM
6.3.1.2
The Film-Tech Forums are designed for various members related to the cinema industry to express their opinions, viewpoints and testimonials on various products, services and events based upon speculation, personal knowledge and factual information through use, therefore all views represented here allow no liability upon the publishers of this web site and the owners of said views assume no liability for any ill will resulting from these postings. The posts made here are for educational as well as entertainment purposes and as such anyone viewing this portion of the website must accept these views as statements of the author of that opinion
and agrees to release the authors from any and all liability.
|
Home
![[Smile]](smile.gif)
Printer-friendly view of this topic