Film-Tech Forum ARCHIVE: Source for replacement DSS100 and DSP100 PSUs

This is topic Source for replacement DSS100 and DSP100 PSUs in forum Digital Cinema Forum at Film-Tech Forum ARCHIVE.

To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=16;t=002927

Posted by Leslie Hartmier (Member # 7053) on 01-24-2017, 02:21 PM:

I noticed in the Datasat thread that someone (Leo?) mentioned using Supermicro modular PSUs in the DSS200, and it led me to wonder:

Anyone know where we might source out replacements for the DSS100 and DSP100 PSUs, and if so, any additional information on them would be keen.

We're not upgrading the devices if we run out, we'd just get rid of them, but it would be foolish to get rid of a device if all that is necessary is to replace PSUs.

That said, it seems that, other than the wee IDE bootstrap drive (and the odd RAID drive) that eventually expires, it's just the PSUs that die, and it would be good to source the things out.

So, any tips on sources and info for the PSUs (and the IDE bootstraps and Dolby-approved HDs, if you have a list of suppliers)?

Thanks!

Leslie

Posted by Martin Murray (Member # 162) on 01-24-2017, 03:18 PM:

Ebay

Posted by Leo Enticknap (Member # 534) on 01-24-2017, 08:33 PM:

I can't tell you anything about the DSS/DSP100 power supply unit, as I don't have any of these servers - sorry. FWIW, the Supermicro PSU model used in the DSS200 is PWB-801-1R. Obviously, another one in that series with the same form factor but a higher rated output would also work.

As for HDDs, if you're registered on dolbycustomer.com and have registered your unit to your account, the list of approved drives is a PDF that you can download from there. However, given that the DSS100/DSP100 is no longer officially supported by Dolby, what do you have to lose by putting non-approved drives in it? If the storage configuration is the same as in a 200, it's a RAID5 and the drives are hot-swapable. So, just as long as you have a spare on hand and are confident to swap it in quickly if the need arises, I can't see any argument against putting cheap, white label desktop grade drives in it. I have one such drive in one of my DSS200s: it's been spinning for 2.5 years now, with zero errors or reallocated sectors.

Very few people would have a 10-year old, 500,000 mile car serviced at a main dealership: to do so would be throwing money away, given the absence of any warranty. IMHO, the same logic applies to a DCP server for which the OEM has ended support.

Posted by Marco Giustini (Member # 4544) on 01-25-2017, 03:29 AM:

quote: Leo Enticknap
I can't see any argument against putting cheap, white label desktop grade drives in it

I do

NAS drives are optimised for RAID. Their firmware will drop the drive quickly if something is wrong on the drive while a desktop-class drive may take its time to try and return the required information.
On a desktop drive a few seconds' delay once in a while is not a big deal. On a Cinema server it may cause the film to stutter.

That said, if the drive is working fine there won't be any difference but manufacturers are using enterprise-graded drives for a good reason.

Posted by Carsten Kurz (Member # 5396) on 01-25-2017, 05:46 AM:

True, but you an always try to buy a RAID optimized drive that is not on the approved list. Since so many people use personal NAS devices now, they are easy to find.

- Carsten

Posted by Marco Giustini (Member # 4544) on 01-25-2017, 05:51 AM:

agree. If not "enterprise" graded, at least a NAS drive, they are not too expensive compared to the desktop line.

Posted by Leo Enticknap (Member # 534) on 01-25-2017, 10:09 AM:

quote: Marco Giustini
On a Cinema server it may cause the film to stutter.

Have you ever seen this actually happen? The volume of data/bandwidth involved is so small, and the number of tasks that a DCP server does is so restricted, that I'd be very surprised if this is a significant risk.

It takes a DSS200 around half an hour to ingest a typical 150GB DCP via the CRU bay. So let's say for argument's sake that it ingests (and therefore is capable of data transfer to and from the RAID) at around 4x the required read speed for playback. So the drive would have to lock up for a long time, and the other drive in the RAID containing the data that is trying to be read, would have to be having a bad day as well, for this to become a significant risk.

Admittedly I only have one desktop grade drive in a RAID of 4, but I have not seen, nor has anyone else who uses that server reported, any picture stutter.

I probably will put a NAS-grade drive in it when that one finally dies (assuming that it does not outlive the server itself, or at least our use of it), just to be on the safe side, but I'm not convinced that this risk is anything above insignificant.

For a server, or any other piece of hardware, that is still covered by a warranty or for which the OEM offers support on condition that you don't go off piste with it, then I understand the need for, and would always use their authorized replacement (and upgrade parts), and those alone; just as if a car is under warranty on condition that it is only serviced at the main dealership, then you have to take it there, and when it comes to the bill, BOHICA. But that does not apply to the DSP100/DSS100 - I believe that Dolby has even crippled the software updates since 4.7.something so that you can't install them on the 100. So, given that these machines in the field are now in the final phase of their service life and maintained aftermarket, I personally wouldn't be inclined to feel that I have to stick to Dolby's approved part list.

Posted by Mike Renlund (Member # 4675) on 01-25-2017, 11:21 AM:

Here are some thoughts I have, speaking very generally as drives do differ across manufacturers:

Enterprise class drives will normally be placed in an array with the redundancy that RAID provides, so if there is an error (bad sector, etc.) the drive will spend a minimal amount of time (seconds) trying to recover knowing that the data is available elsewhere. Because of this, bad sectors and small issues don't result in a degraded array.

Consumer class (desktop) drives are normally the only drive in a system. As such they are designed to go to great lengths to recover data from a bad sector (or other issue). Sometimes taking minutes to do this. During this time the drive may be offline, or not responding. This is where the RAID array gets degraded, or has problems while a drive is working on the recovery.

Exactly how a consumer drive or drives will operate when there is a problem in Digital Cinema playback depends on several factors (bitrate, what content is being called from the drive, how much content was buffered before the error, etc.).

There could be no problem at all, some small stutter that probably won't be reported by an audience, a brief pause, or other manifestation. If it does affect playback it should be short...as the RAID array will degrade and not use that drive.

If the array is degraded multiple times due to a consumer drive being unavailable while trying to recover data, the rebuild process puts extra wear/tear on all the drive(s).

My two cents...

Mike Renlund

Posted by Frank Cox (Member # 6258) on 01-25-2017, 11:22 AM:

I have had that happen, back when I was playing American Sniper a couple of years ago. One of the hard drives in my GDC was apparently bad and there was a section of the movie during the funeral -- about ten seconds that would stutter just like the film was stuck in the gate. Every time.

Replaced the hard drives and the problem was solved.

This was with the drives that came with the server, directly from GDC as far as I know. I replaced them with more drives directly from GDC.

Posted by Mark Gulbrandsen (Member # 72) on 01-25-2017, 11:36 AM:

GDC provides very good HGST drives for the full size servers, but the HGST drives they supply for the SX-3000 external raid box leave a bit to be desired....

Mark

Posted by Frank Cox (Member # 6258) on 01-25-2017, 03:06 PM:

My server is a SX-2001A. I think the hard drives in it are Hitachi, though I could be wrong....

Posted by Steve Guttag (Member # 268) on 01-25-2017, 04:24 PM:

Hitachi = HGST. The HGST badge came about when Western Digital acquired them.

The drives that drove most people nuts were the Seagates from the 2010-2012 servers. They sucked biggly. The big server companies at the time used them and they all migrated to the Hitachi drives.

I have had good luck with the Hitachi, HGST and Western Digital (WD2000FYYZ and WD2003FYYS...both discontinued, natch). If I were to bank on a drive today, it would probably be the HGST line, until proven otherwise.

Posted by Marco Giustini (Member # 4544) on 01-25-2017, 05:15 PM:

quote: Leo Enticknap
Have you ever seen this actually happen? The volume of data/bandwidth involved is so small, and the number of tasks that a DCP server does is so restricted, that I'd be very surprised if this is a significant risk.

Leo,
What Mike said. I didn't say that consumer drivers are SLOW, they are as fast as the enterprise-graded ones.
Just that if there was a read error they would spend a long time trying to recover it - stalling the whole RAID array. A NAS drive will drop out of the RAID immediately, as Mike said.

So, as long as both drives, consumer and enterprise, work fine, happy days. But if anything happens, then you may end up with a stopped/stuttering show when using consumer drives.

Posted by Leslie Hartmier (Member # 7053) on 01-25-2017, 05:34 PM:

Hmm... as I understand it from Gordon, the DSS100s have some fishy setup that prevents using enterprise/NAS/archival hard drives that are not on the Dolby-approved list (he further says that the DSS200s are not restricted in this way), so mostly it's just we'll keep trying to acquire the drives from the list.

I looked at the BIOS of the approved and unapproved drives, and cannot see anything specific that would prevent their use, but we did have a spate where there were a fair number of non-approved drives (installed by the previous owners in...Nashville, I think?) that enjoyed dropping out, and the locations would cry and we'd have to remote in to just rebuild the RAID, so there might be something to the paranoia.

The annoying IDE NAND drives are a lot more of an annoyance (I replaced one recently, and want to find a reliable source for those), but the modular power supplies seem to be the things that like to crap out, and I like to keep extras (also an extra Show Player and Show Store) on site for the location to swap in as necessary, so I'm certainly motivated to source substitutes/replacements.

Leslie

Posted by Stephan Shelley (Member # 8587) on 01-25-2017, 08:14 PM:

The PSU in the DSS-100 is a GIN-6350P. It was a standard supermicro supply. They are available online. The DSP is another story. I don't think they are a standard supply.

Posted by Scott Norwood (Member # 30) on 01-25-2017, 08:58 PM:

Agreed with pretty much everything that others have said. Desktop-grade disks will work fine until they hit a read error, at which point the RAID controller will kick the drive out of the RAID and will probably keep doing so even if you reconstitute the array manually. Essentially, they are fine for temporary use, but not what you would want in a cinema server (or any other real time system) permanently. This has nothing to do with data rate--it has to do with how the RAID controller behaves when it fails a disk, and that almost always involves a period of I/O disruption.

Note that non-Dolby servers use software RAID implementations and not all of those work this way. Linux's md RAID implementation does not have this issue, for example.

Less importantly, enterprise-grade SATA disks have a lower nonrecoverable error rate (typically 1 in 10^15 bits read) than desktop disks (typically 1 in 10^14 bits read). "NAS" disks are normally desktop disks with different firmware and have the higher error rate.

Agreed that Hitachi (owned by Western Digital, but apparently made in different plants) is currently the leader in mechanical disk reliability. The "Ultrastar" line is the one to get for cinema servers.

Posted by Steve Guttag (Member # 268) on 01-26-2017, 06:26 AM:

I do not believe the DSS100 is restricted on drives...particularly with current software for it (4.7). It is only restricted on drive size. Early RAID cards would not support 1TB drives, later (and every one I've tried) do support the 1TB drives. I've never tried a 2TB drive in a DSS100. At present, all of the DSS100s I service are in backup duties.

Posted by Leo Enticknap (Member # 534) on 01-26-2017, 11:14 AM:

quote: Scott Norwood
"NAS" disks are normally desktop disks with different firmware and have the higher error rate.

Given that they're typically 20% more expensive than their desktop-branded equivalents, I find it unimpressive and a bit of a scam if the hardware is no better. I find it hard to believe that the R & D costs on the firmware can account for the difference.

Thanks for everyone's comments. I guess I must have been lucky with the desktop drive I currently have in a DSS200. It was bought as an emergency measure from a local Fry's following a complete failure (of a Dolby OEM-supplied one). I did buy a NAS-grade one shortly afterwards and have kept it in the supplies cabinet as a spare ready to swap in ever since: I didn't see any reason to pull the desktop drive until it actually gave trouble. Given their behavior under partial failure conditions, I think I'll do so the next time I'm in that booth.

Posted by Mark Gulbrandsen (Member # 72) on 01-26-2017, 11:45 AM:

I think it is way more than just firmware Leo. I have found that the WD Red hard drives are about the best out there for serious IT use. HGST IT grade would be the next best and it is what I use in cinema servers because the price vs. performance is just about right. The WD Red drives are way too costly vs. HGST drives to be installing them by the case lot in a cinema application.

Mark

Posted by Scott Norwood (Member # 30) on 01-26-2017, 12:26 PM:

The Western Digital "Red" series is mostly (entirely?) made up of 5400 RPM drives. The performance on these is pretty dismal. You _really_ do not want to mix 5400 RPM and 7200 RPM (10K, 15K, etc.) drives in a single RAID group. This will kill performance for everything. (Actually, you really want every drive in a RAID group to have identical specifications.)

As for desktop drives vs. NAS drives, you can look at the specifications and there really aren't many differences aside from the firmware. The unrecoverable error rate is normally the same for both. Part of the cost difference is probably justified by the (typically) longer warranty on the NAS drives and also because of greater competition and volume in the desktop market.

In the intermediate and long term, it will probably be harder to find enterprise-grade mechanical drives. Data centers are rapidly moving to flash-based storage for performance reasons, as well as reduced physical space, power, and cooling requirements.

Posted by Ioannis Syrogiannis (Member # 3235) on 01-26-2017, 01:27 PM:

I would confirm, dropped frames on the same time-code (spot/sceene) of a movie, due to one of the 4 drives of a DSS200 being problematic. Even more, with the drive (a Seagate Barracuda ES.2) having no alocation sectors at all, reported. I am all for using the same type of disks on a RAID array.
I wouldn't suggest to anyone buying a Seagate product, but I that's just me.

As for the PSUs of the DSP-100, that is a custom Dolby make, it might be repairable by replacing specific parts.

Now, as an hypothesis:
If I had a DSP-100 and I ran out of PSUs, I would consider (until server replacement) to take the PSU backplane out and use a commercial one.
That, due to dimentions, would leave the lid open, but, at least, I wouldn't have a black screen.
I wonder what would have happened to all these dead PSUs if there wasn't the filter on the DSS100...
As for the IDE bootstraps, I 'd go for ebay, just as Martin suggested. I am afraid a re-installment of the OS (content deleted) would be needed.

Other than the above, for the DSS/DSP100 all hardware support has been moved to a company called Venture Manufacturing.
You can contact a Venture representative by email at lrocks(at)venturemfg-usa.com or by phone at 1-510-456-9398. (That is a two and a half years old info.)

Posted by Marco Giustini (Member # 4544) on 01-26-2017, 03:26 PM:

Scott,

The RED's are all 7200rpm - I have some in my NAS. Specifications here.

Leo
That's my understanding. WDC RED's are basically GREEN with a different firmware that make them less prone to be dropped by a RAID controller.
Don't you pay £100 for a "coming home" feature on your car - when it's basically a few lines of code already there on the main computer?

If you want something actually better, then you get the Enterprise (Black) drives. Those are not only 20% dearer!

Posted by Steve Kraus (Member # 476) on 01-29-2017, 11:43 AM:

Several years ago I had freeze / tick playback issues on my DSS200 (running Dolby-approved drives). An ingest was happening at the time so I pulled that and the issue was resolved. No errors were shown but a reboot after the show said one drive had failed. So much for the system properly prioritizing playback data flow over other activity. Or reporting the issue. With another screening following closely, I opted to run that "live" from the transport DCP to minimize use of the degraded RAID.

These were fairly new 1T drives and I still had the old drive complement, untouched, 500GB's, so I put those back in and ran with those until a matching replacement drive could be obtained. Oddly, the alleged bad drive checked out okay on subsequent testing so I have no idea what was going on.

Trying to recall what I did with it but I think it's one I keep in the booth, popping into a CRU box now and then to back up DCP test material.

Posted by Frank Cox (Member # 6258) on 01-29-2017, 11:48 AM:

When I have a degraded array error I get a new set of hard drives and change the whole works. I've done that twice since I got the server in 2011.

Posted by Mark Gulbrandsen (Member # 72) on 01-29-2017, 02:42 PM:

quote: Marco Giustini
The RED's are all 7200rpm

With the gigantic sized sata drives pricing coming down I can see these raid 5's becoming a thing of the past. Throw a couple 4 tb drives in raid 1 and you are done, would have faster read/write speeds and you'd have more safe local storage than you'd probably ever use.

Mark

Posted by Ioannis Syrogiannis (Member # 3235) on 01-29-2017, 02:53 PM:

quote: Steve Kraus
Trying to recall what I did with it but I think it's one I keep in the booth, popping into a CRU box now and then to back up DCP test material.

If you 're backing up from the DSS200, keep in mind that the packages (lacking all DCP elements) will re-ingest only in Dolby servers (DSS100/DSS200/DSS220) and libraries.

You 'll have to extract the generated packages for them to be ingest-able elsewhere.

Posted by Leo Enticknap (Member # 534) on 01-29-2017, 03:04 PM:

It's going to be interesting to see how this goes. My NAS box at home has 2 x 6TB WD Reds in RAID1 configuration in it, which cost around $250 each last year. The total cost of the package (the 2-slot Buffalo NAS box plus the 2 6TB drives as a RAID1, versus a 4-slot box with 2TB drives in a RAID5 configuration) came in at around $200 less than the RAID5 alternative with the same capacity. So even by summer 2016, RAID capacity was cheaper by using a smaller number of drives in a RAID1 configuration than a larger number as a RAID5.

If the 6TB and 8TB SATA drives prove to be as reliable as their smaller predecessors and come down in price to match them in $ per TB within a couple of years, I agree with you that RAID5s of 4-8 drives in "medium sized data" applications such as DCP servers will likely become a thing of the past.

As for SSDs, they're a no brainer for frequently accessed data (e.g. operating system volumes) now, but from the prices I've seen for larger ones, you'd have to save a lot of electricity and gain a big reliability improvement to justify the initial cost in using them for online storage of digital assets of tens of GB or more. At the moment they seem to be like electric cars: in a total ownership and operating cost comparison over a typical lifestyle, not worth it yet.