Basic Topography Questions : Trying to remember the 30K view

I think it would help if you described your workflow a bit more. However...

1: If I just have one raid and one workstation, I only need to set up a DAS via the fibre card in my Mac Pro to the RAID, right?

If you have a single edit station, then yes, simply connect the array to the machine and away you go.

2: IF I were to add a second RAID, I would need both a fibre switch AND a metacontroller?

No exactly. Again, assuming you are talking one workstation and no desire to aggregate the storage into a single volume for capacity or performance issues. With a Mac Pro you can easily install two 4 port FC cards and get two Active/Promise/SataBeast-Boy systems connected for more DAS. Again, assuming one workstation still not direct need for a SAN.

3: IF I added a second workstation, but still just had ONE raid, I would also need a switch and a meta controller?

Depends on how you want your workflow to work and how you plan on doing a render cluster. Assuming that each editor has distinctly defined job roles, you can avoid Xsan and simply remain a DAS environment. Each edit system remains in a silo and data is an island on each machine. You can then use Qmaster's distributed render function to do NFS mounts between the systems and allow for shared resource renders.

4: Can I run the meta-controller on my single mac pro?... If so, would I notice a significant performance impact?

You can... but you run many many risks. A workstation tends to reboot. Rebooting a san takes more that Apple menu > Restart. Also, the nature of running the MDC on a workstation will give you a performance hit. MDCs should be dedicated devices and they should always be deployed with a backup.

Oh, and speaking of backup, make sure you plan for that as well. Raid is not backup. Make sure you implement a solid backup strategy.

5: All of this being said, would these all be seen on the SAN as viable resources for a render farm?

Sure. But there may be a cheaper and easier (although not as fast) way to accomplish it without a SAN. Again, how many systems do you have, how much space (office space) do you have, and what kind of budget are you looking at?

Thanks in advance to anyone who can help. I realize are not specific issues I am looking for a resolution to, but I'm trying to build out a solid SAN for video editing a ** ton of footage. Thanks!!

Xsan provides a number of benefits. If you need these, then Xsan is for you:

• aggregation of multiple independent arrays into a single contiguous volume (huge storage)
• ability to tailor volumes to specific tasks (ingest, frame workflow)
• ability to join multiple workstations into a "storage area network" benefiting from speed and availability of the same data across multiple hosts
• Ability to build render cluster using fiber, not ethernet

Hope this helps. You really need to consider your workflow and your number of users.

Adam - Trigger Studios wrote:
Thanks so much for the detailed answers. There are several possible workflows that I'm considering and so I'm trying to remember all the avenues I can take. Really, I expect more than one editing bay with access to the same data/footage.

Glad to help. If you are expecting multiple edit bays, and you expect content sharing, than a shared file system is the way to go. While I recommend Xsan, I should point out that this is also available. http://www.sonnettech.com/news/pr2010/pr091010_fusionrx1600vfibre.html I have zero experience with it as it just crossed my path and we are planning to investigate the product as it may be a more cost effective solution for small 2 to 3 user environments.

1: DAS just seems like we would outgrow it too fast, and have no room for expansion with data or another workstation.

I would agree unless your build out plans are 12 to 18 months. In which case, migration from DAS to SAN is trivial. But building out a SAN and then growth plans disappear can be an expensive experiment.

2: I understand that you can simply add another FC card and plug in another raid but again, the island thing is probably no good. I realize how we COULD do it, but it seems like bad design.

Agreed, but it goes back to workflow and data sharing. I support a couple of environments in which tasks and projects are rigidly managed by individuals and they have no need for rapid sharing of content. In these cases, a traditional file server allows for the archive of jobs but the isolation of the work project on the workstation.

3: So, if I want two or more workstations to share data on the RAID, it makes sense to design as a SAN. This would of course require a metacontroller+switch. Yes?

Yes. Best practice is to have two switches, and two controllers. Also, while it is still not shipping (which is really unfortunate as I have two environments on hold waiting for it) you should look at http://www.getactivestorage.com/innerpool.php to solve the metadata issue. Qlogic switches are the easiest to configure and when you call them they actually know that Apple products work with their devices.

4: Forgot about the reboots. Definitely want it on a separate machine.

Yes, for sure, we will have long term archival storage. I've seen plenty of raid arrays not rebuild right.

I'm still curious about the render farm. The workstation clusters will see eachother via fibrechannel?

Well, not technically, but yes. What they all "see" is the SAN volume. That is the key, having the shared file system. This allows all the workstations to "see" the same data at the same time. Then with a product like qmaster, you will have a scheduler that will allow multiple workstations to process the same task, targeting the same data, and working toward the same result.

Hope this helps. Good luck with your explorations.