-
Apple – please focus on FCP X stability!
-
[Jeremy Garchow] “If it’s Volume level, though, why do you need a SAN Location? Is it one Volume per machine?”
Well, yes, I presume that’s true. This system uses one write volume per workstation, but it could in theory also be set up with a volume for each production or client. I’m not really sure how it would work in a wider implementation on this system. I don’t really have an issue with it – SAN location is really a tangent to this discussion. My concern is whether there is some optimization that Apple needs to do, specific to this type of configuration, which would improve stability and eliminate the latency or whatever is causing the beach-balling.
– Oliver
Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com -
[Oliver Peters] “SAN location is really a tangent to this discussion. My concern is whether there is some optimization that Apple needs to do, specific to this type of configuration, which would improve stability and eliminate the latency or whatever is causing the beach-balling.”
Aplogies. I can’t find many more people that have any SAN and are running FCPX, so I tend to ask a lot of questions about how people are using their SAN when I do.
I think that as with any SAN that has a dedicated metadata master, that writing small files can throw the SAN for a performance slow down. I think that FCPX’s journaling system, or autosave, or whatever the hell it’s doing when you “make a change” might be the cause of some of those pauses. I try and leave a finder window open to watch the “journaling” happen while I work just to see if that process is what is causing it. It seems to be, but it’s a guess.
Jeremy
-
[Jeremy Garchow] “I think that as with any SAN that has a dedicated metadata master, that writing small files can throw the SAN for a performance slow down. “
Agreed. With FCP 7, the projects on this system reside on the local internal hard drive. When I’ve used Xsan systems, that was also the recommended practice, although I’ve worked with that both ways. Since the events folders only have aliases, I presume on the next project I should run the projects and events folders on the internal as a test. The downside to that is then you end up with render files on the internal drive. Bad design.
Nice of Apple to have not really thought this out very well before tossing the app out into the wild to let users fend for themselves 😉 A little “best practices” guidance would really help (and use only Xsan isn’t it) !
– Oliver
Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com -
[Oliver Peters] “The downside to that is then you end up with render files on the internal drive. Bad design.”
Yes, setting a render file destination would be good….but…
If you are NOT on a Volume level SAN, the rigid FCPX Event/Project structure is actually very convenient as you can simply move the files or load a SAN Location, and everything is exactly how you left it no matter what machine you are on. In that case, it’s a very interesting design.
Perhaps the autosave or journaling system needs to be local to help prevent slow downs.
[Oliver Peters] “Nice of Apple to have not really thought this out very well before tossing the app out into the wild to let users fend for themselves 😉 A little “best practices” guidance would really help (and use only Xsan isn’t it) !”
What would be nice is if you could mount folders on a Volume (SAN or otherwise) that are similar to SAN Locations in that you can put the Project/Event anywhere and not just in the specific root level. This would help with Project/Event Management and allow you to mount/dismount whatever you want without Event Manager X or quitting/moving/relaunching.
When you do work with a SAN that can support FCPX San Locations, it becomes a bit more clear to me that there’s some bigger thinking going on here as it does work really well, and works very differently than just a Volume situation. Also, with metaSAN v5 and their new ProjectStore system, I can now “check in and check out” SAN Locations to prevent other users from mounting each other’s San Locations (this, of course, isn’t a problem with XSan).
What isn’t working, is referenced Event media, but I can’t figure out if that’s because our storage is NTFS or is it’s because metaSAN doesn’t recognize the Aliased file type. I am hoping to setup a test very soon with a vendor and HFS+ formatted storage. This will point to one or the other.
Jeremy
-
Bizarre Oliver. Must be this networked way you’re working because FCP X is extremely solid on a standalone iMac with pegasus raid and BlackMagic Intensity Extreme. I don’t get beach balls, crashes or lost render files. If anything, the way FCP X deals with render files blows me a way. I can render something, then change a setting or two, then change that setting again back to what it was and the render is used again. FCP knows that it has a render file cached that matches the current settings of an element. In any other NLE you’d have to hit undo to get the render back. That’s impressive.
FWIW I’m working with all h264 native content. Often as multi cam. Hardly ever set it to better performance. Usually 1080p24. Turn off background rendering. I keep everything on the Pegasus raid. All files and projects. Gets about 300-500mb/sec.
I’m also running a copy on an old MacPro 1,1 with a 5770 GPU. It renders slower, has a little less playback capability, and doesn’t seem to like the Matrox mini audio out so much (has crackle and static) but overall works pretty well. Everything on it is on a Graid eSata drive. But just with a $50 sonnet sata controller card. Not a sata raid type. Get about 100mb/sec.
I turn off the dropped frames warnings on both because I really could care less about the occasional dropped frame. Not very applicable to editing unless you’re trying to go to tape.
-
[Bret Williams] “because FCP X is extremely solid on a standalone iMac with pegasus raid and BlackMagic Intensity Extreme. I don’t get beach balls, crashes or lost render files”
As I said in the first post, standalone system users probably don’t see a lot of this. At home I’m running a standalone MP and also no big issues. Certainly no beach balls and only the rarest of crashes. I do lose render file links all the time. I’ve now seen this on every system I’ve tested. It could be related to whether you are running media as file copies/imports in the Events folder or aliases.
From what I’ve seen locally, iMacs with Promise RAIDs are a real solid system for X. MPs aren’t. I do question the BMD card though, especially since it’s currently using beta drivers. I’ve had some audio-related issues on two different systems now and the common denominators are BMD, CS6 and FCP X. IoXT is the best with FCP X that I’ve tested (Thunderbolt on an iMac).
[Bret Williams] “then change a setting or two, then change that setting again back to what it was and the render is used again.”
That’s the way it’s supposed to work. Been that way with Avid for years.
[Bret Williams] “FWIW I’m working with all h264 native content. “
Me, never. I always convert to ProResLT or ProRes. In this case, it’s media from a C300, but recorded to PIX240 as ProResHQ.
[Bret Williams] “I turn off the dropped frames warnings on both because I really could care less about the occasional dropped frame. Not very applicable to editing unless you’re trying to go to tape.”
One thing I’ve noticed is that even with the dropped frames warning on, you will frequently see what looks like dropped frames, but the indicator doesn’t say so. Obviously an issue with how the graphics card is addressed.
– Oliver
Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com -
[Jeremy Garchow] “I think that as with any SAN that has a dedicated metadata master, that writing small files can throw the SAN for a performance slow down. I think that FCPX’s journaling system, or autosave, or whatever the hell it’s doing when you “make a change” might be the cause of some of those pauses.”
This.
FCPX Projects and Events are SQLite databases and all the constant I/O with those combined with the I/O for media streaming is a real strain on a typical small video SAN. For high-bandwidth media streaming, you need a wide pipe but not necessarily a lot of spindles. For I/O intensive transactions, you need lots more spindles, and preferably 10K or 15K RPM SAS spindles instead of the 7200 RPM SATA that is common in video environments due to its low cost and high capacity.
When Apple says FCPX is for the next decade, they aren’t kidding- it seems to want SSD-grade storage performance (high bandwidth and high IOPS) to be able to do its I/O. They could work around this pretty easily by allowing the Renders to be sent somewhere other than alongside the Project database files. You can already place media on its own optimized storage by not copying into the Events folder on import.
Best,
Andy -
You might see dropped frames or what appears to be on the computer monitor, but not on the real monitor. Computer monitors are obviously just a point of reference for video. You can watch youtube and you’ll see plenty of judder, but look at the info and it’s not dropping frames either.
That’s great that Avid worked that way. I don’t recall it working that way, but perhaps it was/is. FCP Legend certainly wasn’t that smart. Change anything and then change it back and you’re got to rerender. Unless you go the undo route of course.
Perhaps you should try using the native files. Why would you want to increase the drive space used and complicate the database. Sure it shouldn’t be a problem, but sounds like it is. If you’ve got to convert everything to ProRes, I feel like there isn’t much compelling reason to use X over legacy.
-
[Andrew Richards] “When Apple says FCPX is for the next decade, they aren’t kidding- it seems to want SSD-grade storage performance (high bandwidth and high IOPS) to be able to do its I/O.”
Yikes! That doesn’t sound very promising. You are describing something that has limited functionality for most facilities within a 5-year time frame.
– Oliver
Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com -
[Bret Williams] “Perhaps you should try using the native files. Why would you want to increase the drive space used and complicate the database. Sure it shouldn’t be a problem, but sounds like it is. If you’ve got to convert everything to ProRes, I feel like there isn’t much compelling reason to use X over legacy.”
One reason I convert everything to Prores is that I often edit stuff shot with different format cameras. I find that multimedia plays together much better if you wrap it in prores rather than try to edit say HDV and H264 on the same timeline. Maybe X handles suchlike better.
Reply to this Discussion! Login or Sign Up