Insignificant, compared with...

[mdlbear]

Quick observation:

sbisson quotes from a press release today:

Today, in a significant milestone for scientific grid computing, eight major computing centres successfully completed a challenge to sustain a continuous data flow of 600 megabytes per second (MB/s) on average for 10 days from CERN(1) in Geneva, Switzerland to seven sites in Europe and the US. The total amount of data transmitted during this challenge--500 terabytes--would take about 250 years to download using a typical 512 kilobit per second household broadband connection.

Sounds impressive, doesn't it?

And it's quite true that we're talking about a CD per second, or a half-dozen or so DVDs every minute. But it's still true that storage is getting cheaper much faster than bandwidth: 500TB is only 2000 250GB hard drives, of the sort you can buy down at Fry's for about $125 each. Next year it will be half that many drives, and the price per drive will be about the same.

Now, my measurements show that a hard drive is about 1" by 4" by 6" -- 24 cubic inches. All 500TB will fit in a box 20" by 40" by 60", provided they don't have to be powered up. And if you put them in the back of a minivan, or the cargo hold of a jet plane, of course they wouldn't be. I imagine you could deliver a box like that to each of the seven sites in half a day by courier, or about a day by overnight express.

At work, it takes me over an hour to download a CD image over our T1 line. It only takes half an hour for me to drive home with it. Let's not even think about the bandwidth of a 747 full of SD cards, shall we?

Comments

This doesn't take into account...

[aerowolf.livejournal.com]

This doesn't take into account the time required to hook them back up, the inevitable errors that will creep in when transporting disks long-distance (and you can't just pack them solidly, you need to have some cushioning foam, for protection against bumps and jostles).

As well, after they're all hooked up, the data must still be read so that it may be usefully acted upon. That's a function of the disk controller (I presume you're referencing SerialATA drives there, and I don't know a) how many drives their controllers support, or b) what the throughput is for the bus), and the RPM/throughput of the drives themselves.

Not to mention actually writing all the data to the drives in the first place, disconnecting them, and packing them.

So yes, the time is less to physically move the magnetic or flash-memory substrate that holds the bits, but the time it takes to marshal the data into a usable and useful transfer format, and then demarshal it into the native format that is useful, almost eats up the entire amount of time that was originally saved.

(I've been taking a very 'functional' view of computers of late -- yes, the technology is wonderful and all, but people have to use these things to get work done. The entire time it takes to prepare it is part of the 'TCO' that MS keeps bandying about.)

Re: This doesn't take into account...

[mdlbear]

Well, allow another 50% for padding; there's still plenty of room. Use 350MB drives, say, in RAID5 arrays. Come to think of it, you could save a lot of time if they were packed in 3U rackable modules, blade-style, with a controller in each module. Say, 4000GB per unit. That would be only 125 units, for a total of 375 rack units of height. Since they're shallow, you can fit them back-to-back into six waist-high rolling cabinets.

Disconnect your power and network cables, roll 'em onto the truck, roll 'em off at the other end, and plug 'em in. A couple of hours at each end, tops.

And by the way, TCO is a much less useful measure than ROI.

TCO vs ROI

[aerowolf.livejournal.com]

Disconnected power and network cables also means that you don't have to have duplicate connective cords at both sites, which is a much more efficient way of doing things. Good call. :)

TCO vs ROI: This is also true -- but for some people (notably bean-counters), they want to tighten the belt as much as possible, even if that means a lesser return. (I've seen companies get burned by this mentality before -- it's not pretty when you've spent $135,000 on MS licensing, and it turns around and forces your UNIX admins to learn NT, and then end up having to use UNIX anyway because NT doesn't perform some vital task or doesn't have an easy means to create the connective glue to make the pieces work together.)

[chaoswolf]

Now we take SD cards/USB drives, putting your MP3's & user icon on it (your Kelley Freas one), then sell that as a promo to get people to buy the albums?

[mdlbear]

An SD card or USB drive, even a small one, is probably a lot more expensive than a CD. Takes at least 64MB for a compressed album, if I remember correctly.

On the other hand, nothing says we can't make the CDs bootable, with a small Linux distro on them. I'm thinking Geexbox or Movix (find 'em on Google).

[chaoswolf]

Just a suggestion. It's probably cheaper to do bootables than it is to do full out cds untill we're ready. or what we could do is make the bootable cds available as the "demo" until we get the funds to do the real ones.

[mdlbear]

The "demo" is explicitly a fundraiser. Bootable CDs should be the same price as audio CDs; if anything they might be slightly more expensive because there's more processing involved. Check out prices at, for example, http://www.shortstacks.com/ and note the difference between "duplicated" (burned) and "replicated" (pressed) disks.

The demo will be a limited run (250 or 500); I'm told that it's a good idea to set the price so that you recover costs on sales of about 100. Remember that the dealer takes at least 1/3 of the retail price.

[ashi.livejournal.com]

Demo music CDs with bootable distros on them... there's a nifty idea... and rather fitting for your music. :)

[gfish.livejournal.com]

As a friend likes to say, never underestimate the bandwidth of a semi filled with backup tapes.