On the three studios I've worked in the pass 9 months, I had a total of 5 power outages [:D]

Two of them were due to rain storms in january, which, for some reason, create power outages in Pacific Palisades.

Three of them were because, for some reason, this other studio's power generator went down. The guy had to call the electricity company to figure out where the problem came from.

Anyway- the thing is, it happens. And you usually don't expect it.

As such, some people think about saving their work all the time. But sometime, you just forget. And boom! power failure.

I always say you can never be too cautious.

Realistically, making backups of your stuff is rarely useful. That doesn't prevent me from backing up like crazy.

About a year ago, my project drive stopped mounting. In more than ten years of using computers, this had never happened to me. I didn't have a backup.

Since that day, my project drive is a RAID 1, with an automatic daily backup on an external drive.

Better safe than sorry.

Jerome

UPSes are a bit like insurance against damage to your data due to dirty power; like all insurance, their cost needs to be weighed against the cost of the unexpected loss or corruption of data happening at the worst time.

In theory, modern operating systems are robust when power is yanked unexpectedly (DOS and non-NT-based Windows don't count as modern, by the way -- they barely count as operating systems).

However, in practice, when a computer attempts to recover from being unexpectedly shutdown, it might restore consistency to the file system by deleting anything that seems out of place. That can include all your recent work, if it was only partially saved, or if it was in the process of being updated when the power went. This can actually be hazardous to data that wasn't being directly updated, since consistency can mean: "that directory is in an unknown state, because a file in it was being renamed, so we need to remove the whole thing in order to restore its consistency".

Many file systems have what's called journalling, a technology borrowed from high-availability database systems, which works to recover data that was in transit when the power went. But you might find that its default configuration only protects the structure of the file system, and not its contents, since it can impose a performance penalty.

Note that things like RAID 1 won't help you here, because they're primarily protection against individual hard disks failing, not power to all of them at once.

How does a UPS help? Well, even if it only provides a few minutes of power to the computers which hold your work, that's enough for them to be shut down in an orderly fashion. Even fairly cheap UPSes come with a cable that can be plugged into the computers they're protecting, and this lets them be notified that power has failed. The computer can then perform shutdown actions (as well as send an alert, perhaps) once available battery power drops to a critical level.

The other useful function a UPS provides is to hide brown-outs, where the mains power doesn't fail, but dips momentarily to a level that's too low to let the equipment work right. In my experience, these are more common than complete power failures, and can leave computers in a highly weird, and generally unusable, state if they're not smoothed out by a UPS.

Of course, many computers have UPSes built into them these days anyway; they're called 'laptop batteries', and can lend a false sense of security to those who're used to the consequences of having the protection they offer, protection that isn't available by default in servers and workstations.

Well I suppose we are slightly OT, but I use a UPS here in France where the electricity in my village is notoriously bad.

It's an APC model and quite happily powers all my gear (MacPro, two Mac Minis + external HDs, Apple 23" monitor, Adams monitors, MCU, MC Extender and Big Knob plus other little bits and bobs)

I wouldn't be without it.

Thanks for the post, fcw. Very interesting.

I mentioned the RAID 1 as "one" part of the puzzle. Having a RAID 1 doesn't prevent from having a "real" backup, since if, by mistake, you delete your files, they're deleted on both drives. So, the backup is there to make sure that even if you *do* make a mistake, there will be a backup from "not too long ago" somewhere.

As you said, RAID 1 is cool only in terms of disk reliability, since the possibilities of two disks failing at the same moment is extremely low. The good thing is that if one disk fails, not only will the current data still be there, but the system is also up and running, so you don't even have to worry about it! That's what I call peace of mind.

Timkiel, what part of France are you from?

Jerome

Tinkiel, which model do you have that supports all those machines?

Last time I looked they weren't up to protecting a single 9600 with two drives in it. I guess that's been a while, though... [:)]

Jerome...five outages in nine months?! That's really weird. I've never lived in the Palisades, but I've lived in the area since 1962 - in West LA, Santa Monica, Mar Vista, Culver City, and Sherman Oaks - and it's never been anywhere near that frequent.

fcw, i'm very thankful for your post - this are very valid points. unfortunately actually in europe we encounter an increasing frequency of brownouts, even larger international outages because energy companies are economizing like blazes and this is nothing computers like ...
and you can bet you have just not saved your work in the moment it happens.

we have lately integrated 2 2,5 kVA UPS (corresponding to this model ) and it provides power for more than 20 minutes for a server with a raid, 2 workstations with monitors and several audio- and switching devices.

if an UPS doesn't *pull* a single workstation any more you might want to consider changing the batteries - they don't survive longer than 3 years usually ....
christian

Jerome - near Carcasonne in the southwest.

Nick - I have an APC Smart UPS 1000 1U Rack. Doesn't last very long when the power goes off (about 5 minutes). But it does the job and wasn't overly expensive plus I only had 1U of rack space left - otherwise I might have considered the 1500 or 2200 versions

Cheers

Tim

@Another User said:

if an UPS doesn't *pull* a single workstation any more you might want to consider changing the batteries - they don't survive longer than 3 years usually ....

We've been using the same little 1,500 VA APC unit for eight years now; in that time, we have replaced the battery once. The unit tests itself every few days by deliberately switching to the battery while the mains is up to see if it can take the load, and reports when the battery is becoming too weak to be dependable.

By the way, in case anyone reading this doesn't know, never put laser printers on a UPS; they can have stupendous surges of current consumption that can exceed the output supply limits, which can result in everything on the UPS losing power even when the mains is available.

Thanks Tinkiel.

I actually have my tail between my legs. Yesterday evening we came home, and the first thing that happened when we walked in the door was that the power flickered. [:)]

My main computer was asleep and all the others were off, so the computer didn't even need restarting. But I guess that did make the point.

@Another User said:

You know, I never put monitors on a UPS ...

but it helps if dialog windows need to be answered before the machine is shuting down [;)] and LCDs don't eat too much ....
christian

@fcw said:

You know, I never put monitors on a UPS ...

but it helps if dialog windows need to be answered before the machine is shuting down [;)]

That tends not to be an issue with Linux and Unix servers; ours rarely even have keyboards attached to them, never mind screens.

And while I won't have Windows on any server that we depend on, I do still manage Windows servers for clients remotely via Remote Desktop, which works fine from my laptop, even though it's running Linux too. [:)]

hehe - the server is not the problem, you know, even windows servers can shutdown unattended [;)]
i was talking about audio workstations (mostly in recording mode) - if you cannot answer dialogs here you may loose a complete session ...
christian

@Another User said:

i was talking about audio workstations (mostly in recording mode) - if you cannot answer dialogs here you may loose a complete session ...
christian

Fair enough; I have my DAW set up to save backups every few minutes, and that seems to work okay without being too annoying.

The other (non-DAW) workstations here run pretty much entirely off the network, so hitting them with an axe wouldn't lose much work. Not that that happens too much.

Hi Guys,

If I may, I'd like to jump away from the power issues for a minute…

I'm just starting on this Mini-Farm journey, and have one Intel Mini 1.83 Core Duo hooked up on the network. It's got VI player installed, and I'm using Bidule to host. Currently without an audio or MIDI solution for getting it talking to Logic on my G5, but that's coming soon (with suggestions from you guys, hopefully). At the minute, I haven't installed any instruments to the mini, and am accessing them over the network (they're on an external drive, connected to the G5). This leads me to my main question:

Does anyone use a Network Hard Disk for VSL?
I'm wondering if it's a good idea to hook up a 2TB LaCie to my Gigabit switch, and access it from all the (eventual) machines.

What would be the main issues here?

Or am I still better off swapping out the internal 5400RPM Mini drive with a 7200?
I want to stay clear of FireWire, as I may be using the ProFire ADAT.

I feel like such a newbie again. it sucks.

Peter

Hi Peter:

my advice is for you to switch the internal drive to a 7200RPM. It's really not hard (and at the same time you could upgrade your mini to 2GB... [:)]).

I don't think using a network hard drive storage is a good idea within this setup. Definitely not as fast as having the samples on your main drive.

Plus, the less traffic on the network the better, especially if you are using MIDIOverLan.

jerome

Jerome, what is this based on? My understanding from Jesse White, who sets up these kinds of rigs for lots of composers (on PCs, not Macs) and wrote an article about it in our magazine, is that gigabit ethernet is easily capable of streaming samples all over kingdom come...while simultaneously running FX-Teleport and MIDI. He uses a server machine with RAID, but that's obviously not for bandwidth.

Have you tried gigabit ethernet and found it not to work? Otherwise, frankly, I'd be more inclined to go with Jesse's advice, since he's done it many times.

And without wanting to sound aggressively argumentative, I also have to ask why it's necessary to install a faster internal drive instead of just using a FW drive for samples. That's what I'd do, even if I were streaming eight outputs through a FW interface.

Again, I apologize if I sound like a weenie - that's not my intention; I just want to hear your reasoning.

I know that Jerome has set a few composers up as well. A local drive would always be more reliable than a network drive, right? In a setup with e.g. 10 Mac Minis, you'll probably have one doing Violins, another doing Oboe, etc. Since they're seperated that way anyway, why not use local drives. One less PC/server, one less thing to worry about.

I still think a Mac Mini is a silly choice for VI slaves. I guess if you're scared to death of Windows it's the best way to go. But 98% of the population uses Windows, so there must be a way to make it work... [8-)] Don't get me wrong, I'm a huge Mac fan, but the PC slave you can build for ~$700 blows the Mac Mini out of the water.

Well, the reason for using a server is that all your machines can access it and you only have to install everything once; plus if one machine goes down, you just reach for the next one (and hopefullly you back up your server [:)] ). I don't have one set up, but the idea does make a lot of sense. By the way, my network has never failed, in fact I've never heard of that happening, so I'm not worried about going over it instead of directly to the SATA bus.

As to Windows machines, I have two of them (plus a third that's not working) and it's not like I'm afraid of them - in fact I like them - but they can be a nightmare when something goes wrong. I mean, if you start up from a system CD, Windows wants to replace your whole system drive, programs and all! So for me the idea of my Windows machines being Macs when I need to troubleshoot them is very appealing.

On the other hand I've posted several times that I don't think Mac Minis are a good choice for sample-streaming machines unless you have an extra FireWire interface sitting around. They have no slots and only hold 2GB of RAM, and the built-in digital interface's two outs isn't enough. Plus you can only clock them through TOSlink, which is not an audiophile way of doing things (lightpipes are jittery). So I agree - although I think Mac Pros would make excellent Windows machines; if you need that kind of power, you can't put together the same thing for the price if you buy the equivalent components on the internet. I checked a few months ago.

Just out of interest, what $700 machine would you put together? Motherboard, processor, etc.

may i join your discussion ... IMO there are more *if's* than clear directives ...
mac-mini vs. cheap PC with 2 GB RAM: id say the mac mini can load 50% more samples, the PC is more flexible regarding soundcard and drives ...
network vs. local drive: much depends on the quality of the gibabit switch and how good the server (resp. the storage of the server) allows concurrent file requests - to simply access another host holding data via network would be not the best idea.
firewire-drives: it has turned up drives and audiodevices commonly on firewire are sometimes troublesome, often firewire drives have no sufficient cooling and are dying early when used for sample streaming.

if all involved components are reliable and selected carefully each of the setups above would work of course.
christian

Thanks cm.

How would you set up the server so it does handle concurrent requests? Jesse recommends one of the Windows Server versions (I forget which one - will have to refer to his article in VI [:)] ) - i.e. it's not just another computer sharing its hard drives.