Film-Tech Forum ARCHIVE: 'Standalone' IMS Rack

This is topic 'Standalone' IMS Rack 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=003115

Posted by Frank Anderson IV (Member # 8929) on 09-08-2017, 02:44 PM:

I was curious to know if anyone may know if there is a way to power an IMS outside of the projector?

Perhaps some rack or other piece of equipment may be available that would allow me to do this?

Posted by Dave Macaulay (Member # 813) on 09-08-2017, 03:41 PM:

maybe Dolby or someone else has such a thing, but - put simply - No.

Posted by Leo Enticknap (Member # 534) on 09-08-2017, 04:15 PM:

I do have a dim recollection of seeing a post on F-T mentioning a standalone box into which an IMB or IMS can be slotted, with dual SDI outputs at the back - the purpose of it was to enable an IMB or an IMS to be used with a Series 1 projector. I can't remember any more than that - sorry.

Posted by Carsten Kurz (Member # 5396) on 09-09-2017, 06:03 AM:

http://digicine.com/mediaplayer/

But's not a generic solution.

- Carsten

Posted by Mark Gulbrandsen (Member # 72) on 09-09-2017, 10:00 AM:

Their IMB runs under Windows 10. Thats creepy if you ask me and a big step backwards. I can just see it happening now where the alert pops up in the middle of a show. "Please Wait... Installing Updates". I kinda figured people had learned Windows OS's don't work for cinema servers...

Mark

Posted by Steve Guttag (Member # 268) on 09-09-2017, 11:16 AM:

GDCs new (probably hasn't been released yet) SX1000 series will have its own outboard cage for use in series 1 via SDI but will also slide into a series 2 as a full IMB.

Posted by Stephan Shelley (Member # 8587) on 09-09-2017, 02:54 PM:

Does that mean that GDC is going to discontinue the model that has the internal media block?

Posted by Mark Gulbrandsen (Member # 72) on 09-09-2017, 03:18 PM:

I think the direction of the industry is such that all external servers will be eliminated. If Dolby would get their act together and build a half way decent IMS it would probably happen much faster. GDC is already releasing their third integrated server and the IMS is still a questionably reliable piece after all this time...

Mark

Posted by Steve Guttag (Member # 268) on 09-10-2017, 10:46 AM:

All I can get out of them is that there isn't a set discontinuation date for the SX2001 or SX2000AR.

I'd have to think that they would before too long. I'm sure it is a matter of numbers. How many are sold and how many they still have left.

I can't claim to have had any significant issues on the IMS2000 and it is tons faster than the IMS1000.

I still prefer the separate server model of system. From the GDC line, I only use the SX2001 and SX2000AR. I dislike the SX3000 due to its VERY slow UI and absurd shutdown requirements. I still wish they had made their OS drives mirrored.

Posted by Leo Enticknap (Member # 534) on 09-10-2017, 11:10 AM:

The only issue I (and many of my customers) have with the IMS1000/2000/3000 is the horrible web GUI, based on the fundamentally clunky and unintuitive Doremi software architecture (e.g. Cinelister and Ingest Manager split into two separate applications), and made even worse by superimposing a pull-down tab structure on top of it. I guess this is less of an issue for people who only use these things through a TMS, and therefore, don't have to grapple with the front end directly (ironically, Dolby's TMS software package is one of the nicest I've seen), though. Maybe that is such a high proportion of the customer base for these things that Dolby don't feel it worth their while to develop a decent from end for the few customers who'd be using it directly.

If the Show Manager GUI from the DSS line could be ported over to the IMS line, though, I would have no significant problem with it.

As for the standalone IMS rack/box concept, I'd be surprised if there is any significant market for it. Is it likely that the Series 1 projectors currently in use will outlast the servers they are being used with in significant numbers? Only if that happens and it is no longer possible to buy standalone servers in any shape or form - even good used ones on Ebay - would anyone be likely to contemplate buying an IMS to Series 1 conversion device, IMHO.

Posted by Mark Gulbrandsen (Member # 72) on 09-10-2017, 01:31 PM:

LOL! And people that think the GDC UI is clunky need to try the IMS. Why Dolby didn't just re-write their own very nice UI is just beyond me...
As for SX-2001 I think they could supply them just about indefinitely. There is already a generation 2 motherboard and cooling fan for them and there are so many of those media blocks in the field that they will be producing them for some time to come. It's certainly not what I'd sell to a customer today unless it HAD to match existing equipment.Which in GDC's case is irrelevant.

Mark

Posted by Marcel Birgelen (Member # 6801) on 09-10-2017, 07:23 PM:

Do we really see any future in third-party IMSes, now since every DCI projector vendor is rolling their own and is more or less bundling it for "free" with the projector?

Posted by Mark Gulbrandsen (Member # 72) on 09-10-2017, 09:13 PM:

I think you will see people stick with the same companies they have always done business with. And so far NEC has refrained from forcing any IMS on anyone... The one they sell is strictly optional.

Mark

Posted by Marcel Birgelen (Member # 6801) on 09-11-2017, 04:25 AM:

Personally, I'd like if the projector manufacturers focused primarily on building projectors and would leave the server business to others.

Yes, people tend to stick around with the stuff that has proven to work for them, I can't blame them either. But in case of Barco for example, it's hard to tell customers to throw the perfectly fine Alchemy IMS away and put in a Doremi/Dolby one, which in my opinion has a far inferior UX compared to the Alchemy. You could go for a GDC IMS, which has, in my opinion, about an equal UX, but it's hard to convince anybody to shell out the extra money if it isn't necessary. Even features that were uber-important before, often leave the table at a whim, once the price becomes clear.

I think Doremi sold themselves to Dolby at exactly the right time. Although they did have some non-DCI related business, their business would most likely have died within a few years.

quote: Leo Enticknap
If the Show Manager GUI from the DSS line could be ported over to the IMS line, though, I would have no significant problem with it.

I'd expected Dolby to start integrating both. The Dolby UX was significantly more polished and streamlined, it's hard to believe they just threw that away.

Maybe it was engineered elsewhere or the people that implemented it left Dolby, but else I don't see a reason why not try to merge both the Dolby and Doremi train into one, superior product.

Posted by Steve Guttag (Member # 268) on 09-11-2017, 07:13 AM:

Leo, I was speaking to reliability of the server, not the UI. Yes, the Doremi UI is an abomination. it is made worse by the Web version, if for no other reason than the ill-chosen color scheme. Like anything else, the more you use something, the more you become accustomed to it. So, either way, you'll eventually become proficient at knowing where the things you need are located. The Cinelister UI needs not only the "Show Manager" facelift, it needs under-the-hood work to make it reliably do things in the Ethernet world. It is great you can make macros and all but I can't count how many times it doesn't fire a reliable Ethernet command. I have to have most of them fire a command twice to ensure the other device receives it. I'm dumbfounded that the Cinelister UI still won't let you know what show will start when you press play. That is a bug from version 1 that still isn't fixed and they likely don't get it is a major flaw. How unintuitive for one to have to check/select on the "EDIT" page before pressing play to know what will be running when that happens.

How much better would the GDC UI experience be if they had on the SMS page a button that took your straight to "Content?" Likewise, to take you back to SMS? What is with forcing the way-station that is the "Control Panel?" From a technician/service standpoint, release the "Configuration" section while the show is running. Those things that absolutely can't change mid show, can still be greyed out but things like the Automation should be able to be tweaked mid-show. I suspect that most things within Configuration could be worked on mid-show. And enough of the encrypted logs.

As for the SX2001A...I just recently put one in. The projector hangs from a pipe grid in the auditorium but its sound system (and server) are in the room behind. The projector is used for A/V as well as DCP. By using the SX2001A, I have the sound where I want/need it (including the UI). There are no boot up/shutdown procedures to accommodate either. We incorporated an A/V start and DCP start switches into the automation system so the server doesn't even boot up for days where it is just A/V work (it isn't a conventional theatre). It is an NEC projector so we had DVI and SDI ports available on the "Legacy Board" for A/V use. It just made a whole lot of sense on many levels including CRU ingest.

In the US, the projector company's IMB/IMS is not a freebie. Typically, one can use a 3rd party IMB/IMS for the same or lower cost than a projector company supplied unit. Remember too, the Alchemy only comes with a 2-year warranty while the 3rd party companies issue a 3-year warranty so it is actually cheaper to use the 3rd party, when all things are considered. I forget if GDC still charges for its 4K upgrade, it shouldn't because they would be the only one (Barco doesn't if you buy the ICMP with the projector and Dolby dropped it from the IMS2000/3000).

Posted by Mark Gulbrandsen (Member # 72) on 09-11-2017, 11:09 AM:

No, NO charge for GDC to play 4K any more. Its kind of a standard thing for a long time now. As for ingest and audio being convenient in a separate room, audio is merely a couple more Cat 6 cables coming down from an IMS and there is no degradation in quality. Or if you are the fussy type use real AES cable.

As for ingest if I had a remote projector I would be installing an enterprise storage system, not the PDP-3000. There are many more advantages to using it than just local ingest. Cost wise the two are similar in price to a large server.

Mark

Posted by Steve Guttag (Member # 268) on 09-11-2017, 11:44 AM:

And without the SX3000, I don't have the extra cables nor have to deal with getting the VGA cable to go 80-feet either. And how is that eSATA cable going to deal with the 80-foot run between the Enterprise Storage and the SX3000? And for what? No better performance, loss of SDI and DVI inputs to A/V and increased cost! The SX2001A is cheaper than the SX3000 with Enterprise storage.

Why try to fit every type of foot into the same shoe when there is, presently, such a good variety?

Posted by Mark Gulbrandsen (Member # 72) on 09-11-2017, 04:22 PM:

What 80 foot eSATA cable? The enterprise server goes in the sound rack, data goes to the SX-3000 over CAT 6, there is no PDP-3000 when you have the Enterprise. I would also likely do the HDMI over Cat 6. It's all pretty straight forward.

Mark

Posted by Steve Guttag (Member # 268) on 09-11-2017, 07:23 PM:

Enterprise Storage uses an eSATA...at least every manual I've seen calls it out (just checked the most recent version on the FTP site). In fact, Enterprise Storage doesn't show any Ethernet cable at all (why would it?).

Posted by Leo Enticknap (Member # 534) on 09-12-2017, 09:03 AM:

I'm guessing that what Mark is describing is not the little GDC box containing four 2.5" drives in a RAID5 that connects by eSATA to the SX-3000, but rather a datacenter file or web server type thing that sits in the audio rack, on the same booth LAN as the IMS. You'd mount it as a liveplay source on the GDC, and then play straight from it, over the network. In other words, it's a glorified NAS box.

I would personally be cagey about this. I can think of two situations in which you're going to need a serious amount of DCP storage available (as in, more than 2-3TB).

1 - A multiplex with 10 or more screens, for which a TMS plus a small amount of local storage for each server is surely the most robust solution. You're not having to transfer DCP data over the LAN in real time for playback (with all the reliability issues that invokes, e.g. if a network switch fails during a show...), but you don't need a vast amount of storage in each local RAID, either. 3 x 500GB drives on each local server, plus, say, a 10TB RAID on the TMS would be more than adequate for a typical multiplex situation.

2 - A single screen setup with unusual storage requirements - for example, like my last workplace, which hosted regular festivals that needed 5-10TB of content to be ingested and instantly available for playback for a few days at a time. The best solution there is lots of capacity on the local server. I upgraded the RAID card on a DSS200 to an 8-drive one and put 8 2TB drives in it.

I'm sure that the "Super NAS" approach described by Mark would work, but I can't see any particular advantage to implementing it.

Posted by Sergey Morozov (Member # 9911) on 09-12-2017, 10:23 AM:

As for standalone rack - does anybody know about USL MBI-100 box?

Posted by Ian Freer (Member # 1940) on 09-12-2017, 09:56 PM:

quote: Steve Guttag
Enterprise Storage uses an eSATA...at least every manual I've seen calls it out (just checked the most recent version on the FTP site). In fact, Enterprise Storage doesn't show any Ethernet cable at all (why would it?).

I concur, the GDC Enterprise and Enterprise-Plus units I've installed definitely connect via the same eSATA cable as the PSD did.

Posted by Steve Guttag (Member # 268) on 09-13-2017, 06:13 AM:

Leo,

Mark knows the difference between the PSD (the unit with 3-4 2.5" laptop drives) and the "Enterprise Storage (Plus)" with 3-5 3.5" drives (the "plus" also having a CRU dock). My guess is he is either confused or never put in the Enterprise storage to not know that it only has eSATA out.

GDC does have another box that can be used in conjunction with the Enterprise Storage called the CLA-2000 to allow a stand-alone Enterprise Storage to then transfer over a LAN. I believe the original intent of the CLA is to work with a TMS system, say, oh, the GDC TMS so that one can easily turn it into an LMS without having to think out the system too much. Put in the CLA-2000, an Enterprise Storage Plus with whatever size drives you need and connect that up to your TMS host and you have ingest with content distribution in a canned solution. I don't know if it can make it a NAS storage solution though (I don't know if one can run a movie real-time that way or not).

I think the notion of playing, realtime, off of a NAS via a switch, at this time, is something that will bite you sooner or later. If content is being transferred over that switch as well as your real-time show, you are going to get slow downs (underflows). Network speeds vary all over the place based on what is happening on the network at that time. Now having the NAS merely at a distance from the IMB but with direct connection, that would be a way and really no different than how IMBs work now. The only difference is where the drives reside (in the IMB in a IMS configuration, in an external box, like the SX3000, or in an external box with the OS, like the SX2000AR.

But back to my point earlier, so you can try to make the SX-3000 fit this application, which it doesn't (and he didn't address the VGA monitor output though that too could be another CAT cable with transmitter/receiver) OR, just use the SX-2001A and you have everything you want, where you want it for a LOT less money now, without all of the work arounds.

Leo (not picking on you but you brought it up)...why would you even suggest using small storage locally in your described scenario? A problem my customers have to contend with is CONSTANTLY having to move, delete content to make room for new content. The smaller you make local storage the less flexibility a plex has on moving titles around based on actual customer turn-out. 2TB systems (total capacity) have proven to be a nuisance to my customers whereas 3TB seems to be the tipping point where they have sufficient storage to not worry about constantly deleting old stuff to make room for incoming stuff. Only my festival venues have expressed desires above 6TB total local storage.

Posted by Carsten Kurz (Member # 5396) on 09-13-2017, 06:44 AM:

While Dolby has discontinued the DCP2000 and DCP-2k4, they still ship the ShowVault/IMB solution. From what I see on their support site, the ShowVault now comes with a new mainboard. Has anyone seen one of these new machines in the wild? Together with software 2.8x - is it possible they now support USB3.0 and drives larger than 2TB?

- Carsten

Posted by Steve Guttag (Member # 268) on 09-13-2017, 06:56 AM:

I wished they evolved their IMB such that it was an Ethernet connection to the IMB rather than PCIe. Plus, the HDMI connection is rather wanting too.

Posted by Ian Freer (Member # 1940) on 09-13-2017, 08:57 AM:

quote: Carsten Kurz
Together with software 2.8x - is it possible they now support USB3.0 and drives larger than 2TB?

I've believe they are working on >2TB HDD size with the latest software versions. I don't know if it's mainboard dependent or not, I would have thought it was more a Linux Kernel thing?

I have a couple of new-ish ShowVaults nearby, I'll check the rear panel for any sign of USB3...

Posted by Carsten Kurz (Member # 5396) on 09-13-2017, 09:22 AM:

Would be interesting to know wether 2.8/on a new machine/mainboard uses a more recent kernel? They probably needed do implement some new drivers with the new mainboard anyway. Yes, support >2TB should be doable with a new kernel or other software changes as well.

- Carsten

Posted by Mark Gulbrandsen (Member # 72) on 09-13-2017, 04:46 PM:

quote: Steve Guttag
Mark knows the difference between the PSD (the unit with 3-4 2.5" laptop drives) and the "Enterprise Storage (Plus)"

LOL! Actually I didn't until just now. The first gestation of the Enterprise I saw did have ethernet connection, but it was just a prototype. I agree on the use of eSATA for that. Totally stupid way to go. I would not even consider installing one until it is ethernet compatible!!! That is the dumbest idea for storage I have ever seen, although if you need a shit ton of local storage it's the only way to go. Making it eSATA only is keeping it and this industry in the dark ages.

Mark

Posted by Marcel Birgelen (Member # 6801) on 09-13-2017, 05:37 PM:

Well, eSATA isn't necessarily a bad thing. It's one of the cheapest high-bandwidth interconnection technologies available. It also has far less overhead than anything over Ethernet or even over IP and therefore also a very low latency. All aspects which are great for storage. [Wink]

Obviously, eSATA isn't the most flexible interconnection technology, although there are some eSATA extender products on the market.

But seriously, if anybody starts to talk Enterprise storage, I think about FCAL and iSCSI, not eSATA...

Posted by Steve Guttag (Member # 268) on 09-13-2017, 09:38 PM:

So I'm guessing you haven't installed an Enterprise Storage then, eh?

One of my biggest gripes about eSATA is it is a poorly thought out friction based connector. GDC did use a screw down version, thankfully. I've had miserable luck with eSATA cables. Tripplite cables have been the most reliable for me.

Posted by Mark Gulbrandsen (Member # 72) on 09-13-2017, 09:42 PM:

Yes, well... it kinds keeps it as a local drive storage system... it could certainly be remoted, but with some difficulty... I think GDC really blew it this time. iSCSI might be a little too extravagant an interface. I can't think of a situation needing 10 gig bandwidth in D-Cinema application. You certainly wouldn't want to support more than a couple screens off any content server anyway. I will just stay with using the PDP-3000 for now until some other more common sense option emerges. It's capable of holding four 2tb hard drives. So it's a decent amount of storage. And if some single screen needs more storage you can just swap out the PDP-3000 with another identical unit with different content loaded, reboot the system and have that content available.

Mark

Posted by Marcel Birgelen (Member # 6801) on 09-14-2017, 02:13 AM:

We're now in the realm where 10GE is becoming pretty affordable, especially given the budget of your average multiplex installs. For average 2K content, there is no need yet for 10GE, 1GE will do just fine, but we're slowly moving into a world were we will see content that challenges the 250 MBps limit of DCI. If you're looking at MJPEG2000 content at 4K and 60 fps, you're easily running into 500 MBps+ and beyond. Since doing storage over Ethernet involves all kind of overhead, you don't want to be running this over 1GE. Also, 10GE could give you a major advantage on ingest times over the network, especially when the local drives in the IMSes get replaced by SSDs and those disks are no longer a limiting factor. Also, with shorter ingest times, having smaller local disks, isn't such a problem anymore.

Not that I'm a big fan of centralizing storage for the obvious reasons of reliability, but you can clearly see that this is where it's heading. Modern storage systems are easily capable of pushing out the required bandwidth for 10 screens or more.

Posted by Steve Guttag (Member # 268) on 09-14-2017, 07:42 AM:

Long term, I see fiber is the end-game solution for moving data. It doesn't have the length limit of copper, it is a reliable connection, inherently isolated, and tons of bandwidth.

I agree, long term, it is going to a more centralized storage for content (and even those are going to shrink in size too, with technology. Heck, we have large spinning disc hard drives now to the point I think RAID-5/6/50 solutions will be unnecessary. Why not mirror up a couple of 10GB HDD rather than make 10GB out of several smaller drives? The drive capacity is only going to grow and move more and more to solid-state. But at the other end, projectors are going to turn into just screens so imagine your storage system where it is convenient, yet reliable with a fiber run to the screen.

As far as the PDP, no thanks. I'm not at all jazzed about using 2.5" HDD. They are laptop drives by nature and even the "Enterprise" versions are not as reliable as the standard 3.5" drives. I'd use an Enterprise Plus unit over a PDP every time.

Posted by Leo Enticknap (Member # 534) on 09-14-2017, 09:07 AM:

quote: Steve Guttag
Leo (not picking on you but you brought it up)...why would you even suggest using small storage locally in your described scenario? A problem my customers have to contend with is CONSTANTLY having to move, delete content to make room for new content. The smaller you make local storage the less flexibility a plex has on moving titles around based on actual customer turn-out. 2TB systems (total capacity) have proven to be a nuisance to my customers whereas 3TB seems to be the tipping point where they have sufficient storage to not worry about constantly deleting old stuff to make room for incoming stuff. Only my festival venues have expressed desires above 6TB total local storage.

I suppose that if a 10-plex wants to ingest all the week's features (let's say 12, factoring in a couple of one-off weekend matinees) into all 10 local servers to enable an instantaneous change of screen, then yes, 3TB of local storage would be necessary. Haven't (yet) seen anywhere that does this, though.

Agreed on the likely obsolescence of RAID5 approaches. 10TB enterprise grade drives now have a street price of about $500, so two of those will cost around the same as six 2TB ones (equivalent RAID5 capacity), and save significant power over, say, a typical five years of being spun up.

Posted by Steve Guttag (Member # 268) on 09-14-2017, 10:29 AM:

Leo, please. It isn't just the features, it is all of the previews, ads, and crap that add up on a local server. Yes, the feature are the big ones but a lot of previews add up too (they are the silent drive filler). It has been my experience that 3TB fills up in a 6-plex that wants all current features available on all screens.

As for plexes putting all movies on all servers, We have a couple sites that do that. Some TMS systems do it by design.

Posted by Mark Gulbrandsen (Member # 72) on 09-15-2017, 11:16 AM:

quote: Marcel Birgelen
Modern storage systems are easily capable of pushing out the required bandwidth for 10 screens or more.

I would certainly never place centralized storage on ten screens though. Two screens at most. Ten screens of centralized storage you would want triple redundancy so your entire ten screen is never off line.

And yes, RAID will go away in Cinema servers, likely very soon. There is little point to it and no real need for it any more with the yuge drives that are inexpensively available.

Mark

Posted by Leslie Hartmier (Member # 7053) on 09-15-2017, 02:45 PM:

RAID's arguable primary purpose for cinema servers might be increase of space, but no less important is the redundancy. One drive croaks, no content is lost, and overnight rebuilds returns the system to a state of health. If space was the primary purpose, they would have only gone with RAID 0, which would have been cheaper by a LOT, and far less overhead processing cost.

Using an Ironwolf 10Tb is great, but nothing makes you angrier than to have to reingest 10Tb of stuff because you have lost your drive, and frankly, larger storage past a certain size just means that the location is likely to get in the habit of not doing content housekeeping.

Leslie

Posted by Mark Gulbrandsen (Member # 72) on 09-15-2017, 03:23 PM:

Actually having a 4 TB Raid 0 price wise is almost exactly the same as an equivalent Raid 1 or Raid 5. Drives are so cheap. The money is actually tied up in high end RAID card. Every Raid card out there can do most common Raid configurations through an easy on board menu set up. The only reason for Raid 1 or 5, 6, etc. is redundancy. Raid 0 some times will increase read speed which is really the only reason it even exists. Has anyone ever had a need for Raid 0? Just because you have a redundant Raid also doesn't mean a show isn't going to be interrupted if a raid drive fails during a show. Au contraire... Many Raid cards are either going to kick that bad drive out of the Raid cards automatically go into rebuild if a hot spare is present. Having a redundant RAID only means that once the broken raid goes into a rebuild mode that it will again play back ok and eventually fully recover.

With SAS IT grade drives as reliable as they are today, a 4 or 8 tb SAS drive is likely going to last the four or 5 year duration all by itself and your server will be way way simpler than it is now. Solid State will eventually replace all this electro-mechanical data storage crap anyways.

Mark

Posted by Steve Guttag (Member # 268) on 09-15-2017, 04:44 PM:

Leslie,

My argument is that with the size of the drives that for LMS systems, you can pretty much go to a RAID1 (mirror) and for screen servers again use a RAID 1 for redundancy and just get the size you need (e.g. 2 4TB drives if you want 4TB).

As for RAID performance, the DSS servers will run movies while rebuilding the RAID as well as transfer content. GDC and Doremi grind to a halt if the show is running.

Posted by Mark Gulbrandsen (Member # 72) on 09-15-2017, 06:23 PM:

quote: Steve Guttag
GDC and Doremi grind to a halt if the show is running.

They are both using a version of Intel software Raid. Once you create a new array it doesn't need to rebuild before it can play back. It simply throttles the rebuild process back so it stays out of the "red line" area. But rebuild is still taking place. Most Dell PERC raids also work that way. All initialization and rebuilding is done in the background unless you select for it to happen otherwise.

Mark

Posted by Marcel Birgelen (Member # 6801) on 09-15-2017, 08:43 PM:

Software RAID works well, if the machine you're using it with, does nothing else but storage. If you build a dedicated NAS using some open source or commercial NAS distribution and you're using somewhat recent hardware, you'll most likely even see better performance from it than from a comparable "all-hardware" solution.

The problem is though, a playout machine has all kinds of other things going on and that's what making things challenging. Also, most software RAID implementations suffer the same problem as hardware RAID implementations: They're usually not optimized for throughput, but for reliability.

The danger of random freezes due to a single drive problem is usually larger with software RAID than with hardware RAID. The reason for this is, because the software doesn't have as low level access as the RAID controller. In some cases, a disk which is causing timeouts, causes lockups on the whole SATA bus for example, which is entirely out of reach for the OS, which is handling the RAID layer.

There is also an inherent problem with software RAID1, especially if you're booting from it. The reason is that because when one disk gets dropped, it often doesn't mean it's entirely gone. So, if it was your primary boot disk that got dropped and you reboot the system, there's a chance you're booting from the disk that contains old and sometimes even corrupted data.

Posted by Mark Gulbrandsen (Member # 72) on 09-16-2017, 11:11 AM:

Yes, well both hardware and software raid each have their own pros and cons. Best thing to do is just get rid of it completely in cinema servers.

Mark