Forum Replies Created
-
[Jim Scott] “As far as extensions go my folder has 224 of them, most of which show dates much earlier than my Snow Leopard install. When installing Snow Leopard I did choose to have “settings” moved over from my old system so I assume the extensions came along then. As far as which ones to keep, and which ones to get rid of… I haven’t a clue. Do you?”
There’s no way to know for sure since they may be from other manufacturers. Drivers generally don’t come along when you upgrade. I’d do the archive/install thing and let it go from scratch.
Steve Modica
CTO, Small Tree Communications -
Do you have extensions in /System/Library/Extensions that are older than the snow leopard install date? If so, those may need to be removed/upgraded. You don’t want old drivers laying around.
Steve Modica
CTO, Small Tree Communications -
I think there’s some confusion between block and volume access and file permissions and simultaneous access.
FibreJet and Xan do the same thing. They protect the inode data from being stomped by two guys at the same time. You try to create a file or delete a file while someone else is, and you both splatter the same inode data.
Assuming you have FibreJet and/or Xsan, the inode table is safe, but now you have to worry about the users. UNIX is happy to let many users open and write to the same file! So two users could easily splatter a project.
Some apps will lock a file, some people will use asset management (like FCS) and some people will set up workflows to prevent this type of thing. Permissions can also be used so only one person has access at a time.Fibrejet won’t save you from this type of thing. (Image editor 1 finishing his work and editor 2 mounting the volume and changing it in an undesirable way. This is a workflow issue)
Steve
Steve Modica
CTO, Small Tree Communications -
Hey Bob,
I’m the guy on the righthttps://www.motorcityrock.com/bands/inasense/inasense.html
Steve Modica
CTO, Small Tree Communications -
Steve Modica
January 3, 2011 at 2:01 pm in reply to: Final Share and 1+hour sequences stuttering question for Walter BiscardiAre you seeing the generic drop message or the “slow disk” drop message?
Since you are mixing formats, I would render the timeline and try again. Assuming you are only running one stream on the timeline, this will show that when FCP doesn’t have to transcode on the fly (to handle the mixed formats) it will work.
We’ve been cataloging all of these little quirks.
Walter did us the kindness of sending us a long video sequence that was very complex and we found many things like this:
Still images on the timeline
mixed codecs on the timeline
reversed sequences
sped up sequences
various audio formats (like mpeg)Depending on the speed of the system, these things can cause drops reliably, or just make them more likely to occur. If the system is an imac without jumbo frames, they are *much* more likely to occur. In fact, we use ours as a test bench since it’s more sensitive.
With our stuff, we make sure that from spindle to FCP, all of the data shows up on time. What final cut does with the data once it gets it can be a mystery 🙂
Steve Modica
CTO, Small Tree Communications -
Does anyone have insight on Avid/Protools and why they continue to require block storage?
In the olden days, a realtime application would want direct block (raw block) access so they could write to the end of the volume. Sort of like tape. SGI had a realtime filesystem like this. People would call because they would delete huge projects yet have no more space. That’s because we never used blocks in the middle. We always wrote at the end, so deleting things in the middle didn’t help. I can understand why people did this because “best fit” algorithms will slow things down and make them “non-realtime”.
These days, with audio being so tiny and disks/filesystems being so fast, I don’t see the point. We can do HD uncompressed video off of disks with normal filesystems, so why not audio?
They don’t need to do that anymore.Steve Modica
CTO, Small Tree Communications -
[Lorenz Redlefsen] “(I don’t think it’s fair to compare the per-port cost of an el cheapo 10/100/1000 switch with an enterprise-class 10G-capable switch, with redundant power, cooling, etc. It *is* true that there aren’t any el cheapo 10G-capable switches today, so 10G does not compare favorably when you compare “cheapest-to-cheapest”.)”
I think for most of the people on this forum, the enterprise stuff is cost they don’t want to incur. (For example, redundant power and fans). In most cases, the switch is off at night and they never use layer 3. They want really fast layer 2. One reason we carry edge-core is they are very inexpensive, but use a nice broadcom chipset that can do flow control correctly (FYI, the 10GbaseT vendors we’re aware of won’t do flow control correctly with a negotiated 1Gb link. I’d love to hear if someone out there can!)
Steve Modica
CTO, Small Tree Communications -
Steve Modica
January 3, 2011 at 1:44 pm in reply to: Tony Silanskas problem, and “doing it yourself”I’ve noticed that the qmaster daemon is pretty chatty. When we’re looking at problem tcpdumps and kernel.log files, we can usually spot qmaster right away.
Steve Modica
CTO, Small Tree Communications -
We were almost called Little Tree. But one of the engineers thought it sounded too “cute” so we went with “Small Tree” 🙂
The current imacs all use the broadcom 5764 chip and it does not support jumbo frames.
Our own testing showed that it made the system more prone to drops. For example, an imac will drop a frame when a still image is on the timeline (Pro Res). It can be rendered and then it won’t drop anymore. A Mac Pro shoots right through.Steve Modica
CTO, Small Tree Communications -
This is a bug in the GUI. They are probably set. You can see with ifconfig. It’s an old bug. It just doesn’t show the change.
Steve Modica
CTO, Small Tree Communications