How Many CDs Can You Burn at Once?

kfs27 asks: "In an attempt to help a professor of mine record and duplicate his lectures. I have been asked to put together a CD duplicating box. Commercial products seem to be very expensive and I figured a PC with some SCSI160 Cards (HW or SW Raid maybe), SCSI Burners and a 15K RPM drive (size not an issue) could do the job for cheaper. But the question is, how many CDs can you burn at once of 30 minutes, mono audio. 10 at a time would be excellent I think. More of course better. Cost is not a huge issue, as long as it's less than Commercial Duplicators, it's more of an experiment, but must be stable and easy to operate (I'd be willing to script up a frontend)."

Comments

RAM Disk, not Hard Drives

[Zack]

With the prices of RAM being as low as they are, you might want to consider building a ram disk to store the data that's going to get burned. That way you don't have to worry about the speed of the hard drives, the ram will always be faster.

We have one box here with 4 SCSI burners in it with a 700 meg ram drive. Everything works wonderfully in it.

Re:RAM Disk, not Hard Drives

[jonistron]

hmmm, I really like this idea about RAMDISK. But don't forget duplication services like http://www.ctexinc.com/ or http://www.cd-rom-replication.com/ . Take your total cost of your design and weigh it against using a company like these. 10,000 cd's should run about $4,000.00 US

Well at a glance..

[jonistron]

[(max bandwidth to drive)*(n disks)]/[(max bandwidth cd-r drive)*(n cdr's)] > 1

Plextor Replex Tower

[haplo21112]

The Plextor Replex tower would be a much better way to go...it and its software are made for this sort of thing and are reasonablely cheap...about the same cost as what you are suggesting, after a quick calculation in my head..

Umm, depends.

[polymath69]

Factors are going to include the speed of the drives holding the .iso images, the number of CD burners per SCSI bus, and the speed that you are trying to burn the disks at (eg. 4x). The idea of using a RAM disk to eliminate the hard drive variable has some merit, but unless you've got a gig of RAM or so, you'll only be able to burn one image at a time onto those multiple disks.

But here's an unrelated thought. You say these CDs are to contain 30 minute mono lectures. A CD will hold 60 minutes easily. Why not put two lectures per CD and save on your overhead in loading and unloading the disks?

You could take it even further by recording one lecture in the left audio channel, and another in the right, to fit four lectures per disk. It might be worth considering.

Re:Umm, depends.

[polymath69]

The AC is right; it wouldn't make sense to mount the ISO itself, as it's the image that's of use to the burner.

But here's an even better idea. Skip the ramdisk. Instead, create a bunch of named pipes like np1..np10. Then start up a bunch of cdrecord instances, each reading from one of the named pipes. They should all block until they fill their input buffers.

Next, use tee to copy the ISO image from the hard disk into each of the pipes.

$ tee np1 np2 np3 np4 np5 np6 np7 np8 np9 < file.iso > np10

That reads the ISO from the disk only once, and so should have the same performance advantages as setting it up in a tremendously large RAM disk (the contents of which might get swapped out anyway.)

Note that the final named pipe was redirected to, rather than named as an argument to tee. That avoids the need to write an extra copy out to /dev/null.

Will it work? You'd have to try it...

PCI bus is your bottleneck...

[stienman]

The only bottleneck you have then is your PCI bus.

You'll probably be fine with the 10 drives and one HD as long as:
1) You use a ramdisk
2) You make sure each burner has at least 2MB of buffer

With the 2MB buffer, fast scsi, ram disk and DMA you should run into no problems even with 24 or 32 speed burners. You'd be better off, of course, with a faster/wider PCI bus.

Integrate a robotic loading/unloading system, and 24x drives - you'll get 10 cds every two minutes. Your class of fifty can get their CD on the way out the door. It may be more cost effective to get twice as many drives that run at half the speed.

-Adam

Re:PCI bus is your bottleneck...

[hamjudo]

Don't make this too hard. The PCI bus on a cheap motherboard is 32bits at 33Mhz, or a peak bandwidth of 132Mbytes/second. A 1X CD-ROM is 44.1Khz * 2 channels * 16 bits (2 bytes) or 176Kbytes/second. If the PCI bus were perfect, it could drive a single 748X CD writer, or ten 74X CDWriters. It's not perfect, but you'll easily get more than half of the max bandwidth (assuming modern PCI cards). Since 32X is the fastest CDwriter available today, you can easily drive 10 of them with quite a margin.

If they're new CDwriters, they'll have protection against making coasters, so the penalty for running too many CDwriters, is they'll slow down.

If you're good with metal working tools, you can make a double wide case that can hold 10 drives. Plug a bunch of IDE controller cards into the PCI bus. Note: IDE cables are limited to 18 inches. So you have to do a little design work, before you start cutting metal.

If you're not into metal working, just take a few cheap PC's, max them out with CDwriters and network them with 100base-T ethernet cards. A little glue, and it will be like one big machine. This design can be expanded to hundreds of drives.

I'm assuming student labor, so you won't need a robot disk changer.

Let the students burn...

[martyb]

Let the students burn the CDs themselves. Just set up a server (ala napster), tell the students to download the lectures. Then, if the students actually want to burn them to CD, they're free to do so (set upa FAQ, if need be).

Why CDs?

[Jerf]

Why CDs? 30 minutes of mono audio, encoded in 32Kb/s MP3, is (30 * 60 * 32 / 8) = 7200 KB (with the last "/ 8" to get us to kiloBYTES instead of kiloBITS). There are MUCH better codecs then MP3 for this at this bitrate, I just use MP3 as a convenient and easily obtained example. Record the lecture, convert to (lossily-compressed-audio-format-of-your-choice), and load it on the web.

At the end of the semester, give each student ONE CD with the entire course on it!

Nowadays, if your student can use a CD, they can play an MP3. And even a 7MB download is doable over a modem connection. (And you might cut that down to perhaps 1 or 2 MB or less by using a codec designed to do voice-only, but you'll probably have to pay for it.)

Plenty. Is there a pile of Kenny G. . .

[human bean]

That I can light off? :)

don't use a RAM disk

[MSG]

First, I have to say that I agree with the comments already posted suggesting that you just compress the audio and make it available for download. It's much more efficient. Especially when semester end comes and students want to review data from several lectures, not swapping disks frequently will be more convenient.

I'd mod them up if I didn't have to say that RAM disks are a bad idea. If you simply add the RAM to the system, then the OS can cache the data in the most efficient manner possible. As long as you have the RAM to cache the image, the OS shouldn't be reading it from the drive constantly. Using a RAM disk is actually harmful to system performance, because the OS may not be able to cache disk sectors that are frequently needed. Statically allocating the RAM only works if you have more information about disk use than your OS, which you almost certainly do not.

What's the bottleneck?

[Spamalamadingdong]

If all the CDs are burning at once, you have to do one read and many writes. Speed of the hard disk is not a factor. Speed of the bus and drive interfaces might be.

If a CD holds 660 MB and holds 1 hour of audio, that's a data rate of 11 MB/minute. Burning at 24x, that's 264 MB/minute. Bandwidth of a 64-bit wide PCI bus at 66 MHz is 528 MB/second, some 120 times the requirement of the single CD drive. It would appear that one could burn 10 or 20 CDs at a time at 24x and have plenty of bus bandwidth left over (so long as you were burning in parallel).

I'm not qualified to judge the architectural features which might create other bottlenecks, but neither the hard drive nor the machine bus appear to be a difficulty.

Practical problems.

[billcopc]

Burning multiple discs at once does sound nice at first, but there is one thing you seem to have overlooked : bus contention. No matter how many hyperfast SCSI cards you put in there, they will all share the same PCI bus and they will all compete over whose data stream is most important. This leads to reduced total throughput and greatly increased latency.

One thing you could try instead is to just use a bunch of older P2-300's with IDE burners and stream the audio files from a fast NFS or SMB server. Burning at 8x requires about 1400 KB/sec, so good ole 100base-T could serve 4-5 clients without a hiccup. Throw in 3-4 nics and you could have yourself a burner-cluster for very cheap.

Key points highlighted

[autocracy]

• Buy enough RAM to hold the entire image of whatever you're burning - but don't make a RAM disk to store it in. The OS will do just fine caching it (my twin 18 gig SCSI drives that I got on the cheap evidence this - my computer barely reaches for them after the first hour of uptime. REALLY fast!).
• SCSI burners wanted. This might even come out costing less because of the fact that an IDE card can only take 4 devices, while you'll be able to push several times that on SCSI. IDE will suffice for the HD if you're only burning one image at a time though.
• Avoid theoretical numbers. Worst case scenario all the way. Assume that even if you start all your burners at the same time, they'll have drifted by the end of the cycle. And as you do more and more reloads, I can see the timings differing. This won't affect reads of the image (it'll be cached), but writes will hurt. Get a SCSI card and wide enough PCI bus (shoot for 64/66) to take it. Sure, theoretical measurements say you'll hack it with a 32/33, but keep in mind that others things run on that bus besides just what you plug into the slots - and that's assuming 100% efficiency.
• SCSI stressed again. Bus mastering will help your CPU SSSSOOOO much. And yeah, I can't vouch it'd be that great on the ide -> scsi deal - but it's cached in RAM and RAM -> SCSI will help the proc.
• Hardware you'll want: 1 Gig of RAM (why skimp? A full CD is 700 meaning that you'll probably buy 768 (multiple of 256 - derr!), so shoot for the gig to play with. Any CPU P-III or better will do, maybe even a P-II. I'd just go pick myself up a decent Athlon though. As for the Mobo - Dual Athlon model. No, you won't use that second proc, but you will use the onboard Ultra160. If you can get one with dual channels - bonus! This is because you'll be capable of handling 30 drives.

Consideration: You'll have to have something to change out the discs. Set up with 10 drives, you should be able to EASILY do this for ~$2000. It's up to you to figure out way to change the discs automatically, though! And good luk on the case :)

More information Needed ... IDE DRIVES?

[fwc]

There's a lot of information missing from the posters query.

First of all, if these MUST be in standard CD-audio format, then the answer to the question about how many disks you can burn of 30 minutes of audio in a given time can be calculated by dividing 30 minutes by the speed of the reader (say 15x), and then adding a minute or two for lead-in lead-out, toc, loading, etc. In this case, a 15x drive should be able to burn a 30 min CD in about 3-4 mins. A single drive should be able to turn out around 15-20 an hour.

The poster did say he wanted to do this on the cheap. The bandwidth bottleneck in a PC environment will most likely be the PCI bus. Even with two IDE drives on an IDE chain, you should be able to keep up with the burning at 15x (150MB/min per drive). If I was going to do this on the cheap, I'd get me a used Pentium-II/Celeron class machine, or possibly a higher end pentium machine, get 4 IDE chains in it, and load it up with 6 CDR drives. Total cost should be under $1000, assuming you use linux or freebsd or similar. ($600 for drives, $50 for controller card, $350 for used machine). You may need to add a little for memory expansion, as I think the idea of a Ramdisk (300Megish) would be good, but memory is cheap (512 total MB should be sufficient). If you need more drives, add another machine. If you find that the machine can't keep up with this many, drop one or two and put them on a second machine.

If these are for delivery to students which aren't at the lecture, or for review, perhaps the best thing would be to not focus on bulk duplication, but instead to figure out an on-demand system. What I mean is that if a student WANTS the lecture, then they can visit a computer at a specific location, select the date of the lecture, insert media and wait 5 mins for it to spit it out. That would be *Really* cheap (linux box w/CDR and suitably sized hard drive).

Do the math!

[Rick the Red]

He doesn't want to do this once, he wants to do it many times a year, for each of the professor's lectures. If he had "orders over 500" then buying in bulk would make sense; but if he only needs 30-50 copies of each lecture (my guess) then he doesn't want to make them one at a time (but ten at at time would be OK) and he doesn't want/need to buy 500 from a service. If he's making 30 copies of 20 lectures that's 600 CDs, not 50, so it's more like $4/CD if the equipment costs him $2000. But your way would cost $0.89x500=$445/batch (not the $125 you claimed -- your math is flawed somewhere) x 20 batches = $8900 (plus throwing away 9000+ unneeded CDs). Clearly the $2000 homemade box is cheaper for this application, and if he needs more than 30 copies of each lecture it's even cheaper.

Of course, another way around this would be to put CD burners on 10 existing PCs in a lab somewhere. Not as convenient, but workable and cheaper (assuming they can get time to use that lab :-)

Re:Do the math!

[joshsisk]

You said:

your way would cost $0.89x500=$445/batch (not the $125 you claimed -- your math is flawed somewhere)

I said:

If he's building a system for $2000+ to make 50 cds when he could just pay $125 bucks instead, he's crazy.

I think your reading is flawed somewhere, I specifically wrote FIFTY, not 500. (The 500 number, which you took from elsewhere in my comment, was an example of the pricing curve of CD-R services. Note the first paragraph where I say you can get 30-50 made for $2.50 each.)

Again, it still seems cheaper to just get the CDs made as you need them. You pay $125 for each batch (assumming a batch of 50), never have to worry about bad burns, don't have to have someone spend their time "scripting a frontend" or any of that business. Also, you never have to worry about your hardware going bad, or anything really.

. If he's making 30 copies of 20 lectures that's 600 CDs, not 50, so it's more like $4/CD if the equipment costs him $2000.

And furnacecd (and others, I assume) will make cd-r batches as small as ONE cd-r for $2.50 each, with the price dropping the more you make. If your way gets the price down to $4 a CD, it's still cheaper to use a service- $2.50 each at the most. Of course, over several years, using your own equipment will become cheaper- but that's not including time for maintenence, replacing hardware if it becomes needed, setup and actually doing the work.

It also means an investment of hardware and capitol for something they might decide wasn't the best way of distributing the lectures in the first place. If, after one $125 outlay through a service, they realized that the students didn't even use the CDs, they could just not make another batch. No $2000 and all the time wasted.

I'd recommend getting the first batch made through a service for this reason even if they plan to build the equipment. That way they can see if the students even want this. I can easily see the professor asking for a show of hands "who listened to my lectures on the cd?" and only one or two students raising their hand.

Another thing to consider is the space taken for this equipment, and the hassle of requisitioning the hardware. I have worked in the University environment, and in my experience you can do small funding outlays (under $500) with little or no problem... But have to do proposals and seek bids for more expensive items.

Incidentally, I agree with others on here that the best way to distribute lectures is via the web- why even bother making CDs in the first place? Just let them download them, or listen in the library/their home pc.

Disk Changer

[Anonymous Coward]

If the creation of the disks isn't time sensative, you might want to consider some sort of CDRW coupled to a disk changer, so that you don't have to sit there the entire time waiting for it to spit out disks. At 6x write speed 30 minutes of audio is written in 5 minutes, so that would be about 4 hours for fifty copies. Not too bad if you don't have to sit there feeding the drives.

Alternatives to what you want.

[Linuxathome]

Are you just burning audio or are you also including powerpoint slides or images? I'm also a student, and the lectures we have (1 hour lectures at a time) are recorded as wav files and then encoded as mp3 files at a 16K bitrate (trust me, I've listened to them and they work fine for lecture audio) and at that bitrate, the techs managed to squeeze 1 hour in 5 MB's (well, they also squeezed the sample rate down during recording too). That's the size of a regular music track at 192K bitrate, 44.1KHz samplerate--in other words, just the perfect size to publish online and download. Add less than 1 MB for the powerpoint file and you have enough info for most of the students to skip class and just listen to audio while flipping through the slides at home. Let me emphasize again that this is already implemented at my medical school and lots of students skip class due to this feature. ;-)

Ok, my question is: why spend the money on CD duplicators? I think it's more worthwhile to spend it on a computer station with all necessary drives for all available media that the students use. You can even turn it into a webserver if it has fast internet access. That way, all the lectures will be on this station and the students would only need to go to it and pop in their zip disk, jaz disk, cd-r or even better, a cd-rw, and then be able to copy what lectures they want. So, I think rather than spend your time trying to build the cd-duplicator, spend your time on writing the software/program that is running on the station that will allow the student to easily choose what they want and then instantly hit the "Burn" or "Copy" button and copy it to their media. In my view, this station is a much better use of your time.

Actually, if you wanted to make it a truly killer app, then instead of copying the mp3's and the powerpoint files separately, have them integrated with, say, a macromedia program that the students can run independently (without the need of either a mp3 player or even powerpoint) and it'll automatically play the audio and show the slides cued to the audio (no need for the students to guess which slide the prof is on).

But then again, I could be totally offtopic and your reasons behind building this cd-copying system far outweighs my suggestion. Anyway, these are just my thoughts.

Use caution when burning CD's

[selectspec]

Remember what happened to those Texas College students burning those logs? Don't build your CD burning pile too high. Stay clear when you ignite it. Be careful! Children's Barney CDs tend to ignite quickly. Don't burn only Brittney Spears. Don't leave out your David Hasslehoff and Shatner CD's. Be creative with your bad artists.