Andrew Richards
Forum Replies Created
-
Make it stop!
Best,
Andy -
Here we go!
Best,
Andy -
[Jeremy Garchow] “Man, events/projects on raid0? Yeeps.”
No riskier than having it all on a single disk. Remember to backup!
Best,
Andy -
[Oliver Peters] “That poses a render concern, although the single internal seems to playback ProResHQ render files just fine. I’m going to see if this owner is interested in putting a couple of SSDs into drive bays 3 & 4 of the Mac Pro. Software RAID then as RAID-0 and only use them for FCP X Events and Projects.”
Unfortunately, that is probably the best workaround for now. I submitted a feedback a while ago asking for the ability to set separate render storage. The type of storage good at databases is very different from the type of storage best suited to storing and streaming media.
Best,
Andy -
[Walter Soyka] “The point is this: on the PC side, I know that as long as someone who has money wants a workstation, someone who wants to make some money will build one and sell it to them. Apple thinks different.”
Yeah, that was my point- We the Internet all assume Apple will get bored and take its ball and go home, while everyone else would take a real hit from dropping products they sell thousands of. They killed the Xserve because of low sales, so now any Mac that doesn’t sell in seven figure quantities is a FUD magnet.
I’m really looking forward to finding out what has really been going on inside One Infinite Loop come Monday.
Best,
Andy -
[Bernhard Grininger] “Disadvatages:”
Another one:
Working from Disk Images carries a non-trivial overhead for the system that will reduce your performance ceiling in FCPX to the tune of fewer streams or grater UI lag on larger projects.
Best,
Andy -
[Oliver Peters] “2 chassis with 16x2TB drives in each.”
That should still be quicker than a Pegasus for small I/O since it has more spindles. Did you say you do not see these issues when working off a Pegasus?
Best,
Andy -
[Oliver Peters] “It’s 64TB partitioned into 5 volumes – 4 workstations plus FC Server controlling the 5th. Each station has write authority to one partition and all others mounted in read-only mode. During most of this session time, I’ve been the only operator accessing the system. No particular ongoing processes occurring with FC Server, except what it might routinely be doing in the background.”
Probably not an HFS+ shortcoming then. Assuming that is a 32 disk array, I would expect it to perform better at small I/O than a Pegasus due to the higher spindle count. If you can take the same work to an iMac and a Pegasus and it doesn’t lag on you, then I’m stumped.
Best,
Andy -
[Oliver Peters] “The finder shows the write volume as Mac OS Extended.”
Yep, that’s HFS+.
[Oliver Peters] “Why would this be any different than any other locally-connected external drive, like a Promise or even an internal set of RAID-0 drives?”
If other SAN users are reading (or writing, if this array is sliced or partitioned) from the same array while you are working, it would impact the array’s ability to deal with the heavy IOPS writes associated with the FCPX database files. If you are the only one using the array, then it is essentially no different from local-attached storage and that probably rules out HFS+ as the culprit if you don’t see these issues with similar local-attached storage. It really depends on the architecture of this particular SAN.
Best,
Andy -
[Oliver Peters] “I seem to remember that Fast/Liquid was in a constant save condition AND it had true background rendering. Not bad for a 32-bit Windows app that had none of these issues, as I recall.
It might be that part of the problem with X is that it’s actually writing several databases for the Events and Projects. Not just a single project file. One thing I noticed was that when I was adding a lot of favorites (which included writing the selection to a smart collection in parallel) was that it was quite easy for me, as the operator, to outrun the application.”
Another issue might lie with HFS+ (I didn’t see if you specified what type of filesystem your client’s FibreJet SAN is built upon, I assume it is HFS+). It is limited in its ability to do more than one I/O action at a time (as I understand it), so that can also be a limiting factor, maybe even the most significant limiting factor. I wonder if you had an Xsan filesystem on the same hardware if my hypothesis about IOPS would go out the window because the filesystem can multithread I/O actions when properly configured…
Best,
Andy