Manifold Recording - Studio Tech Thread

[Piyono]

Are you talking 48TB of usable, arrayed disk space or just the total raw capacity of the drives?

I'm wondering now about a couple of things:

i) Throughput: Is this server dedicated to audio or will it handle video, as well? If so how many simultaneous HD streams can it manage whilst still running dozens of audio tracks?

ii) Backup: In your usage scenario projections do you envision any one project making use of the bulk of your capacity? A band could easily walk out with an album or two's worth of sessions on a single hard drive, but wow will you secure 48TB of data if need be?


fH

[Clueless]

Originally Posted by Piyono

Are you talking 48TB of usable, arrayed disk space or just the total raw capacity of the drives?

Raw capacity.

I'm wondering now about a couple of things:

i) Throughput: Is this server dedicated to audio or will it handle video, as well? If so how many simultaneous HD streams can it manage whilst still running dozens of audio tracks?

You can obviously do the math, but 64 tracks of 96K audio is 20MB/sec, which is far less than a 3Gb/sec HD video stream, and is easily done with a single RE4 drive. HD video streams do require special handling, and we haven't started test that yet. We'll likely use these drives as storage for editable video, not for capture and not for archive.

ii) Backup: In your usage scenario projections do you envision any one project making use of the bulk of your capacity? A band could easily walk out with an album or two's worth of sessions on a single hard drive, but wow will you secure 48TB of data if need be?

I really don't expect a single session to exceed 1-2TB for all their audio and editable video, so no. Even if I'm wrong by a factor of 10x, that's still less than half our capacity, and therefore not "the bulk". As to bands walking out with disk drives...an SNMP alert will tell us the moment that a drive is removed, and it's all 911 from there. When we've lost utility power to the site, we get a call from our fiber provider within 10 minutes that they cannot see our fiber box on their network. Any service loss in between those two extremes (such as unplugging a server or a network cable) can also be noticed, and we can take appropriate action.

[Piyono]

Originally Posted by Clueless

Raw capacity.

You can obviously do the math, but 64 tracks of 96K audio is 20MB/sec, which is far less than a 3Gb/sec HD video stream, and is easily done with a single RE4 drive. HD video streams do require special handling, and we haven't started test that yet. We'll likely use these drives as storage for editable video, not for capture and not for archive.

I really don't expect a single session to exceed 1-2TB for all their audio and editable video, so no. Even if I'm wrong by a factor of 10x, that's still less than half our capacity, and therefore not "the bulk".

It's the math that got me wondering. I was researching storage for a DAW just over a year ago. When it came to capacity I imagined the most outrageous recording scenarios my mind could invent, then doubled what I thought I'd need, and the figure I came up with was something like 6 terabytes of on-line storage. I figured I wouldn't be working on more than a few projects at a time and that completed projects would be offloaded to other devices. 48 Terabytes of storage seems, well.. a lot— for just audio.

Bandwidth, too. I mean, like you said, any old drive can do 20Mb/s without breaking a sweat, and most any single one of today's 'performance' drives can handle big sessions with tons of edits and fades and so forth. So an array with throughput like yours seems a bit, well... fast— for just audio.

I can't (yet) conceive of how I'd use that kind of capacity and bandwidth in a music studio. That's why I wondered out loud about your projected session size and bandwidth requirements.

Originally Posted by Clueless

As to bands walking out with disk drives...an SNMP alert will tell us the moment that a drive is removed, and it's all 911 from there. When we've lost utility power to the site, we get a call from our fiber provider within 10 minutes that they cannot see our fiber box on their network. Any service loss in between those two extremes (such as unplugging a server or a network cable) can also be noticed, and we can take appropriate action.

Holy miscommunication, Batman!
My gosh, I wasn't talking about larceny! The notion hadn't even crossed my mind.

What I meant was that, from a backup perspective, several albums' worth of sessions can easily be contained on a single hard drive, so backing up something like that is easy, whereas backing up 48TB might prove more difficult if ever the need arose.
Sorry for the confusion!

fH

[Clueless]

Originally Posted by Piyono

What I meant was that, from a backup perspective, several albums' worth of sessions can easily be contained on a single hard drive, so backing up something like that is easy, whereas backing up 48TB might prove more difficult if ever the need arose.
Sorry for the confusion!

fH

We'll back up session-by-session, not array by array, so no problem. We regularly save off ~100GB in less than 15 minutes.

[Clueless]

I was hoping to post the caterpillar->butterfly transformation of either the Pergola or the Annex control room this week, but alas the scale of the work again exceeded estimates. The Pergola team spent the entire week just making the common rafters and the acoustic team spent the entire week making the cloud. The photos of the exciting transformation will have to wait until next week...

[mastermix]

And I was worried that I was going to have to "come down" from the construction thread! Love this geek stuff, keep it coming!

[Piyono]

Micahel,

I've been contemplating data security lately, and methods by which studios, especially studios working on major-label projects might keep releases from leaking early.

How do you manage network access to the media SAN server? Is it controlled by some form of machine- or user-authentication?

Piyono

[Clueless]

Originally Posted by Piyono

Micahel,

I've been contemplating data security lately, and methods by which studios, especially studios working on major-label projects might keep releases from leaking early.

How do you manage network access to the media SAN server? Is it controlled by some form of machine- or user-authentication?

Piyono

We have pretty strong network security rules: VLANs at the switch level and iptables mediating network/machine combos. We deny by default and authorize only what's explicitly allowed. The Linux servers also have pretty strong user authentication. Of course, any security system could always be better...

[Piyono]

Originally Posted by Clueless

We have pretty strong network security rules: ...Of course, any security system could always be better...

Sure, well that's the classic refrain in the security world, right? There's no 'perfect' security, just what's acceptable in a given context.

What I always find amazing is how many studio operators (nay, people in general) have not the foggiest notions of security or data integrity.

That said, I can't imagine you'll have too many ill-intentioned, security-guru 1337 hackers running about the place, so your measures will probably suffice.




Piyono

[lukehatpadl]

Originally Posted by Clueless

We have pretty strong network security rules: VLANs at the switch level and iptables mediating network/machine combos. We deny by default and authorize only what's explicitly allowed. The Linux servers also have pretty strong user authentication. Of course, any security system could always be better...

Yeah, I hope you're not using LDAP ;-)

[Clueless]

Good morning, all! We have some tech news to report...

We've been using ProTools HD Native at 96K in both rooms for a few months now, and we are getting a pretty good handle on its capabilities (or lack thereof). A recent session with 55 tracks on a single socket MacPro (4 cores / 8 virtual cores) with only half the cores assigned to ProTools used less than 5% of the CPU, less than 20% of disk I/O capabilities, but 90 of the 96 available voices.

In other words, the voice limit set by ProTools is about four times too low for a half of a half of a modern day Mac.

As I understand it, playing back 55 tracks uses 55 voices, which by itself is no biggie. But if we want to bus, on average, one aux send and one stereo send per track, that would add another 165 voices! If we want those various aux sends going to 8 different aux busses and the stereo sends to 4 stereo group busses, that adds another 16 voices. Suddenly we're way past 200 voices, easily within the range of what the Mac could handle by itself, but well beyond the artificial limit of HD Native at 96K.

Thankfully, we have consoles. In the API room, our 64 channels have two inputs and our aux, group, stereo, and surround busses all have inserts, for a total of 128 + 32-ish input voices. Each channel can feed a direct out, 10 aux busess, 24 track busses, 3 stereo busses, and a surround bus, for a total of nearly 50 output voices, so a total of 3200 output voices. And then these aux busses and stereo busses have their own folddown matricies, for an additional set of voices. So in ProTools terms, the API console is about 35 times more powerful than ProTools HD Native when it comes to I/O.

In the Harrison room, we our 96 channels have two inputs, for a total of 192 input voices. (We can source from more than 500 distinct digital patch points, but those sources must be routed into the console before they represent a "voice".) Once inside the console, we have a direct out, 16 aux busses, 48 track busses, 8 group busses, and four stereo busses for a total of 80 * 96 minus the 8 busses that cannot route to themselves = 7672 output voices. So in ProTools terms, the Harrison console is about 80 times more powerful than ProTools HD Native.

In both rooms we have an Xdubber stem recorder (or two), which can record and play back 64 tracks in floating point. Thus, we can make a drum stem, a background vocal stem, a wall-of-guitars stem, etc., thereby actually *recovering* voices for ProTools as the mix gets bigger and deeper. When we mix in the API room, we can source ProTools from one fader and the Xdubber from the other: 64 + 64 inputs all day. In the Harrison room, we can have fader-per-channel on 48+48 inputs or do more creative things to use the B inputs for natural summing (such as source + efx return and controlling efx level based on efx send rather than fader return), allowing us to use more of the 64 channels of ProTools or more of the 64 channels of stems, whichever way the session goes.

[Clueless]

For some, the selection of microphones is a technical subject, so I'm letting you all know what we have selected for the studio. The page is here.

[CeretoneAudio]

Nice list!

I think there is a typo - you have listed josephson e42s - should be e22s

[Clueless]

Originally Posted by CeretoneAudio

Nice list!

I think there is a typo - you have listed josephson e42s - should be e22s

Nice catch! I also just finished posting a gear list.