New port is fiber channel [Archive]

mcl

May 30, 2003, 04:52 PM

Originally posted by MarksEvilTwin
Whats so weird or useless about having a RAID system at home? Drives can fail spontaneously anytime, and yeah, if you keep good backups thats great, but we cant always be ready. My dad's HD randomly died last fall and he lost all these photos he took over vacation about a month earlier, because he hadn't had the chance to back it up at all, also some work documents. He now has two drives hooked up in a RAID so he's got these two drives mirroring eachother.

I think RAIDs are a great idea, and i dont think they should ONLY be used in pro environments. If you are willing to shell out the little bit extra for the second drive, its definitely worth it.

Mark

A FC-AL RAID enclosure will run you around $10k. That right there makes it abnormal for home use.

arn

May 30, 2003, 05:40 PM

Originally posted by MarksEvilTwin
Whats so weird or useless about having a RAID system at home?

As another posted said... we're not talking about just RAID... Mac OS X supports RAID to a limited degree in software...

We're talking

FibreChannel up to 2.52 terabytes RAID (http://www.apple.com/xserve/raid/) - which would presumably be the reason to put FibreChannel in a Mac product (by Apple)

arn

timbloom

May 30, 2003, 06:33 PM

I could see a perfectly good reason for them, although I agree it is unlikely to happen in the powermac.
PowerMacs are made for creative professionals, right? They need massive and redundant storage just like the companies running xserves for databases and such. If I had enough money and needed a large RAID, I would prefer not to have the xServe.

anthonymoody

May 30, 2003, 06:34 PM

If this is true it's great news for HD editing (which really requires an array). Right now, to connect a nice array to a PM you have to buy the $500 fiber card. But if they duild it in, you wouldn't have to. I agree that the usefulness of this is limited, but for those who have the need, it's cool.

TM

Sedulous

May 30, 2003, 06:39 PM

I don't know if MacBidouille is saying that this is what they think some of the connectors are or if they know they are fiber connectors.

Steradian

May 30, 2003, 07:02 PM

Hmm I am having doubts about the accuracy of these rumors, I would like them to be real. However I don't think that they are correct. FibreChannel, common do you really need that...? Hmm maybe they are making a new level of consumer, Pro Pro-Sumer

rog

May 30, 2003, 07:05 PM

Rincewind42

May 30, 2003, 07:15 PM

Originally posted by anthonymoody
If this is true it's great news for HD editing (which really requires an array). Right now, to connect a nice array to a PM you have to buy the $500 fiber card. But if they duild it in, you wouldn't have to. I agree that the usefulness of this is limited, but for those who have the need, it's cool.

But for those of us without the need that is another $500 (maybe less, but probably not less than $300) added cost per computer. I would rather the people who need it buy the $500 PCI card than those of us who don't pay more.

KershMan

May 30, 2003, 07:19 PM

I think it is quite possible that a "new" high-end PowerMac line could have fiber-channel drives. Sun makes workstation level machines (Sun Blade 2000) that has fiber-channel drives.

This would give the PowerMac the punch of high-end SGI and Sun workstations. A 64-bit OS and fiber-channel drives would give a PowerMac a significant punch for video professionals. Think about the RAM you can put into a 64-bit OS machine and the speed of the drives. You would have incredible speeds rendering digital video.

mohaukachi

May 30, 2003, 07:27 PM

Originally posted by rog
Uh folks, this is all from the site that claimed to have benchmarks of the 970 using a soon to be released new version of Bryce that is far into development. Guess what? Corel just announced that Bryce version 5 is it as far as the mac goes. Everything on the MB site is just totally and completely bogus, and specifically, news about Bryce is the final nail in the coffin in regards to their earlier claimed benchmarks.

as excited an into a WWDC release of the 970... this really puts me down. wonderful... now we'll have to rename the french rumors "freedom rumors . it sure dosent help the 2k ive been saving for a system thats buring a hole in my pocket.

:confused:

Mac Messenger

May 30, 2003, 07:38 PM

The fiber Rumor is true.......(It seems unlikely that Apple would build-in FibreChannel into PowerMac motherboards) <- This is wrong!

Longey Nowze

May 30, 2003, 07:44 PM

why are all/most the 970 rumours coming from one site??

weird i guess.

thank you
MaT

Longey Nowze

May 30, 2003, 07:47 PM

Originally posted by rog
Corel just announced that Bryce version 5 is it as far as the mac goes.

would you post a link please?

thank you
MaT

arn

May 30, 2003, 07:48 PM

Originally posted by Mac Messenger
The fiber Rumor is true.......(It seems unlikely that Apple would build-in FibreChannel into PowerMac motherboards) <- This is wrong!

also a possibility...

I've clarified that this is my personal opinion in the article.

arn

centauratlas

May 30, 2003, 07:50 PM

>Why would you want to have a RAID system hooked up to a desktop <

I can think of several things that will impact this:
1. It will get cheaper as time goes on.
2. "640K should be enough for everyone" - even if the quotation isn't true, 15 years (shoot, 5 years ago) ago people figured 31-bits of address space "would be enough for everyone", who would need it at home? 2GB of RAM is enough. Now we are at the point where it is NOT enough with iMovie, iDVD, iPhoto, Panther etc.

I think the point is that in the future, you will see mass storage come down even cheaper than now (obvious) where it will add perhaps $50 to the cost of the computer to have a RAID system inside.

Eventually you see the server technologies (like DDR) migrate into the personal machines. For example:
1. Hard drives (remember when it was a tape drive, then a floppy on the Apple ][), then 400K floppies, the 800K floppies (wow lots of space) etc. Then the HD and NO floppy, just CDs. Now DVDs
2. Ditto for separate VRAM and graphics processors.
3. Ethernet on every machine (before that Appletalk)

Apple is generally the leader (dropping the floppy for just a CD), adding a mouse and windows, Firewire, SCSI, built in ethernet, Appletalk etc) in applying technology for the personal computer. Apple needs this type of thing: more memory address (hence the 970), eventually RAID (for reliability and speed), DDR (hence the 970).

Whether it is truly here now is another question, but RAID will be an option for the desktop and eventually not be an option, just be there.

And 20 years from now we'll be saying "who thought 63-bit addressing would be enough for everything? We need a 128 bit processor to handle more memory."

bones

May 30, 2003, 08:15 PM

Originally posted by Longey Nowze
would you post a link please?

thank you
MaT

http://www.google.com/search?q=bryce+mac

http://maccentral.macworld.com/news/2003/05/30/corel/

MikeH

May 30, 2003, 08:19 PM

Although I'd like a lot of the recent rumors to be true I can't help but think a lot are just wishfull thinking by Mac fans with no grasp of economics.

Fibre channel's a case in point. Lets face it, how many people are connected to a Gigabit network let alone need fibre channel? I don't know anyone, at work it's all 100mbs and at home I've got a 10mbs hub - which is fine and can several computers over the internet at once - via a router (two PC's at once still get pings of 25-35 on Counterstrike).

Fibre channel, although cool, would be a waste of resources for 99% of people, so why not leave it out and let the minority that need it buy afterwards - there's no point paying for something most of us will never need.

I don't think this rumor is true and it's just building certain people (fanatics and fantasists) up for a fall.

noverflow

May 30, 2003, 08:22 PM

What we have here is the xserve mothboard

the xserve has as of now 2 fw800 and one fw400
just like they claim that this 970 has

also the xserve has B-9 (RS-232
this is probably an other un known port.

also it would be smart of apple to build in fibre into these. because that will free up a pci slot, speed up the raid, and make the xserve a better buy.

The fibre pci cards you can get for the xserve are dual port/channel.... so that would be one of the 2/3 unknown port bundles.

ffakr

May 30, 2003, 08:35 PM

I would SERIOUSLY doubt Apple is putting FC on the motherboards....
... but, Apple put 1000Bt on the boards back when Gbit PC NICs were about $600 a pop. I couldn't believe that either (after it happened!!!). BTW, who uses 1000Bt Ethernet now? ... a couple years after Apple put it in the machines? Prices are just starting to become reasonable for the 1000Bt switches but they are still way expen$ive, and you better have quality TP cable, and a nice RF free enviorn.

Personally, here's my take on all these rumors (if anyone cares ;-) )

1. MacBid is pulling a lot of this out of their butts... or someone they are talking to is pulling this out of their kiesters.

2. MacBid is probably fairly accurate regardless. The first thing I thought when Apple delayed WWDC was "The 970". I, however, wouldn't be surprised if IBM debuted the chip faster than 1.4-1.8 though. I've heard their yeilds are good, and IBM Germany did pre-announce a 2.5 GHz Blade. Why would they do that if 2.5 was way down the road... like 2004?

3. Apple is just wacky enough to put FC on the motherboard, BUT if they do expect this to be a 64bit Unix workstation. Don't expect this to be marketed to the pro-sumer. :-( That's bad IMHO. I like the iMac, but I want a tower for my 20" monitor... not an all in one.

.... but what do I know, I'm just a stupid Ffakr.

JrdBeau

May 30, 2003, 08:46 PM

An X-Station??? Oh, you mean like this???

http://users.rcn.com/crfriend/museum/machines/IBM-X130.html

Maybe we should call it something else...

-J

bahram

May 30, 2003, 08:48 PM

"Uh folks, this is all from the site that claimed to have benchmarks of the 970 using a soon to be released new version of Bryce that is far into development. Guess what? Corel just announced that Bryce version 5 is it as far as the mac goes. Everything on the MB site is just totally and completely bogus, and specifically, news about Bryce is the final nail in the coffin in regards to their earlier claimed benchmarks"

I could not c the refrance of Bryce 6 on MB:

http://www.macbidouille.com/niouzcontenu.php?date=2003-05-05#5440

but did find it on barefeats who apprently had bench as well

http://www.barefeats.com/pentium4.html

:confused:

BillGates

May 30, 2003, 08:51 PM

Originally posted by MikeH
Lets face it, how many people are connected to a Gigabit network let alone need fibre channel? I don't know anyone, at work it's all 100mbs and at home I've got a 10mbs hub - which is fine and can several computers over the internet at once - via a router (two PC's at once still get pings of 25-35 on Counterstrike).

Fibre channel, although cool, would be a waste of resources for 99% of people, so why not leave it out and let the minority that need it buy afterwards - there's no point paying for something most of us will never need.

Hmmm. If this logic board is for real we're obviously not talking about an iMac. When talking about moving a lot of data we're not talking latency.

A pro computer would benefit from having FiberChannel and Gigabit Ethernet. All of our pro Macs that have gigabit Ethernet built in and several that don't are attached to gigabit switch(s). FibreChannel attaches our arrays to the servers, servers are attached to gigabit switch(s).

In may industries moving data fast is necessary to stay competitive. It would be nice to attach all our workstations to the arrays as SAN via FibreChannel. Maybe Apple has more in store for the xRAID. Maybe Apple plans to make SAN cheap and easy ... xSAN?

Whose to say that FibreChannel (copper version) needs to cost all that much. How much do you think people paid for that first built in Gigabit Ethernet port?

wizard

May 30, 2003, 08:57 PM

"I think the point is that in the future, you will see mass storage come down even cheaper than now (obvious) where it will add perhaps $50 to the cost of the computer to have a RAID system inside

Raid storage is already in this range if your running Linux, are willing to use software RAID, and aare satisfied by the capacity of older drives in the 20 to 50 GB range. In fact I ran RAID on my Linux system for awhile. Up until I could make better use of the drive configured as real disk space.

Whether it is truly here now is another question, but RAID will be an option for the desktop and eventually not be an option, just be there.

I completely agree! As with the mouse, it is difficult to buy any PC without some sort of pointing device. A machine with built in Raid technology will have a marketing advantage.

And 20 years from now we'll be saying "who thought 63-bit addressing would be enough for everything? We need a 128 bit processor to handle more memory.

Well this I may differ on a bit. I think the wave of the future is loosely coupled multiprocessor systems. Its amayzing what SETI and some of the others have accomplished in the crud manner that they have started out.

Thanks
dave

rog

May 30, 2003, 08:58 PM

maccentral.macworld.com/news/2003/05/30/corel/ (http://maccentral.macworld.com/news/2003/05/30/corel/)

Above is the link for the Corel cancels Bryce story as requested.

Flynnstone

May 30, 2003, 09:00 PM

Perhaps Fibre channel on a Power Mac is possible.
But the real doubt in my mind is that there appears to be no ethernet port ! I can't believe a PowerMac or any other computer would come without an ethernet port. :(

andyduncan

May 30, 2003, 09:01 PM

One thing nobody has mentioned is the possibility of these being optical firewire ports. IIRC the 1394b spec allows for up to 3.2gigabits/sec over a few hundred meters of glass fiber, or short runs of copper. This is pure, unadulterated speculation of course. Has anyone even sampled optical firewire controllers?

BillGates

May 30, 2003, 09:04 PM

arn

May 30, 2003, 09:14 PM

depending on the cost of the machines in total...

another possibility is a high-end PowerMac (970, FibreChannel), while keeping the G4's available at current pricings.

arn

Kid Red

May 30, 2003, 09:29 PM

Is this 'fibre' channel spelling a European thing from the translation or the purposed Apple spelling? It's just confusing me when I look at it. I keep reading 'fee-bray'.

Also, it seems like a race to stack up as much 970 info as we can before anything is actually released. Macdoobie is either sleeping with Stevie's European mistress or has the crack pipe working overtime with the shear number of rumors lately.

soggywulf

May 30, 2003, 09:55 PM

Originally posted by arn
another possibility is a high-end PowerMac (970, FibreChannel), while keeping the G4's available at current pricings.

Well, I hope that doesn't happen. The 970 and the fast bus are needed immediately at all tower price points, to be even close to competitive with PC's in price/perf. The existing machines cannot compete at all, especially if they stay the same price (there's no sales now, and there won't be any later either).

AidenShaw

May 30, 2003, 10:02 PM

Originally posted by Kid Red
Is this 'fibre' channel spelling a European thing from the translation or the purposed Apple spelling?

No, "Fibre Channel" is the official name.

Look at http://www.qlogic.com/products/fc_san_hostadapers.asp to see lots of "fibres"....

macnews

May 30, 2003, 10:21 PM

With all these rumors coming from MB I would tend to think they might not be true - esp the 970 not to mention fibre channel.

It might make sense if these are new, high end, xserve motherboards. Apple is trying to make in roads to the high end server market and perhaps they have had some requests for this. I can't imagine they would put it in tower models, and I think we can safely say we will not see these in Imacs. If they do go in powermacs I hope they don't jack the price up too much, not something I would need or want - unless it was a desktop server maybe!

If I can offer any sort of hope that the MB rumors about the 970 are true let me offer a little rumor I have heard. A collegue of mine just got back from an Apple training (at Apple). During one session a guest speaker told the class (it was a small class) to buy Apple stock becuase some BIG changes are coming at WWDC. For what it is worth, I have no reason to doubt my collegue and while no mention was made of the 970 the guest made it sound like this would be more than just something like "you will be soo impressed with Panther" type of hint.

QCassidy352

May 30, 2003, 11:10 PM

Originally posted by Kid Red
Is this 'fibre' channel spelling a European thing from the translation or the purposed Apple spelling? It's just confusing me when I look at it. I keep reading 'fee-bray'.

Also, it seems like a race to stack up as much 970 info as we can before anything is actually released. Macdoobie is either sleeping with Stevie's European mistress or has the crack pipe working overtime with the shear number of rumors lately.

well put. funniest thing I've read in a while.

I agree with the person who said these are x-serve mother boards. We all just assumed powermac... but that was us jumping to the conclusion that we wanted, and was not based in any fact.

dguisinger

May 31, 2003, 12:32 AM

There is the possibility that this is both a PowerMac & Xserve motherboard.

Little has been given on the chipset. What if Apple has built FibreChannel into the southbridge to reduce costs of design? Remember the XServe & PowerMac now use the same chipset which was originally designed for XServe.

I could see Apple integrating FibreChannel and a second gigabit ethernet port into the chipset. And the motherboards are so close in design, if they wanted to, they may be able to design a universal board with maybe a daugher board for the expansion cards so towers can contain more slots. I dont know, thats just speculation.....however reducing overhead by relying on a single pro motherboard would be a good decision.

ces1965

May 31, 2003, 12:56 AM

There were some rumors about system-level clustering a while back. Is gigabit ethernet good enough for that? If not, would fiber channel do it? This is something any power/creative user could take advantage of. Buy four macs and cluster. Anyone have any comment on this?

illumin8

May 31, 2003, 02:15 AM

Originally posted by mcl
...short-haul, high-speed optical alternative to SCSI (and notorious for driver problems on pretty much every platform on which it's been deployed. Sun's one of the the longest-running platforms using FC-AL, and their drivers STILL spontaneously LIP, and don't interop well with non-Sun GBICs.) Bzzt. Thanks for playing. You obviously don't know jack ***** about fibre channel. All of the major Unix players use fibre channel for high capacity storage, and I assure you it works quite well. This includes Sun, HP, IBM, Fujitsu, SGI, and anyone else worth a damn.

By the way, fibre channel now supports 2gb a second transfer rate, and if by "short-haul" you mean several hundred meters then I guess that would be short.

Get with the program. Fibre channel is superior to all flavors of SCSI, except perhaps SCSI-320, in speed, and the ability to have up to 127 devices on a single fibre channel loop is a plus.

SunRiser

May 31, 2003, 04:43 AM

i want to add a few remarks about this fibre channel stuff from MacBidouille...

Being a Profesionnal Video Editor, this FC rumors just add the lacked piece to the pict picture actually at Apple...

First, they started with Final Cut Pro then bought Shake and DVD Studio Pro. FCP and Shake can deal with HD pictures but need a big and FAST storage... SCSI can hardly make it. Even if you want to work with SD (regular PAL or NTSC) and multiple real-time uncompressed streams, yo uwill need a FC stream as well...

Now, all the high-end editing stations are FB-Raid based (like Symphony from Avid).

Apple is very into getting back the video market. They introduced FCP 4 and Shake 3 during NAB '03 but lack a fast computer to use them (the dual 1.4 are ok but still lag in speed in front of PC solutions). Moreover, all the raid options on MacOS X (ATTO and stuffs) are NOT releable under MacOS X... Just see all the problems we have at work with those, ATTO claiming it's Apple's fault...

So what do they need to offer to make the video ppl happy? a fast computer with a fast and storage that work well with MacOS X.

Now the rumors parts: the PPC 970 should make look the Dual 1.4 like a calculator and now the motherboard should have a FB connector in it... See the link? ;)

Of course, it's my personal opinion and i have no insider infos (except was once beta tester for FCP 2 and 3 ) But it seems very logical for me afterall :D

dethl

May 31, 2003, 05:20 AM

Just wondering, but if Apple wanted to include onboard RAID, wouldn't it be cheaper to put whatever ATA(insert your favorite ATA speed here) RAID hardware onto the logic board rather than require someone to chuck out $500 to get a special fibrechannel raid card?

I don't see a personal need for this, but I can see why my sis (who wants to go into professional video editing/directing) would need such.

I would much rather see this as a built-to-order option rather than having it rammed down my throat (not to mention my pocketbook).

Mac Messenger

May 31, 2003, 05:26 AM

Like I said before. The Fibre it true. The digital is true. Let there be no more wondering. What is the next subject?

moki

May 31, 2003, 05:38 AM

Originally posted by Mac Messenger
Like I said before. The Fibre it true. The digital is true. Let there be no more wondering. What is the next subject?

Is that a toslink in your pocket, or are ya just glad to see me? :)

Sun Baked

May 31, 2003, 05:48 AM

With everybody swallowing all this Fibre, I'm surprised we haven't heard more screaming and yelling that this rumor is now clogging up the rumor mill (for some reason the mill looks like a steaming toilet to me right now).

visor

May 31, 2003, 07:11 AM

Originally posted by MarksEvilTwin
Whats so weird or useless about having a RAID system at home?
Mark

You're absolutely right on that one. In fact, i have a very simple Raid1 setup in my little linux fileserver. A cheap controller and 2 HD's is all it took for that. It's not fast, but it definately is failsafe if one HD dies from age.

However, the little linux server is locked up in the closet, makeing to much noise in the workroom.

soggywulf

May 31, 2003, 07:15 AM

If on-board Fibre Channel doesn't cost too much, then fine. If it does, then IMO it's a terrible idea for the towers.

Bottom line is that Apple needs to keep the price points the same on the 970 towers. Even if the prices stay the same and we get dual 1.8s on the high end, we will -still- lag behind PCs in terms of price/performance. I don't expect that the Mac will ever reach parity with PCs in this matter, but it would be kinda nice to stay somewhere remotely in the same league.

visor

May 31, 2003, 07:31 AM

It makes definately a lot of sense - use it as docking station link to a decent computer etc.

Now c'mon guys, you really amaze me. Of couse the fibre stuff doesn't go into a desktop by default. Not even apple would be crazy enough for that right now.

It goes into the Servers, where it makes perfect sense. Those Apple Raid cases belong in an apple raid rack, stacked upon each other, locked away in a sound proof room, thats all there is to that.

so the good news is: Powermacs may still get 5.1 surround sound ;)

withnail21

May 31, 2003, 07:45 AM

I have to agree with those who think that Fibre Channel for a PowerMac (as the current range stands) is highly unlikely. However, it does make perfect sense for an XStation-type machine.

Think about it.

Apple is making some serious roads into high-end video at the moment. They have a great editor in Final Cut and probably the best compositing in the world with Shake. Right now, the hardware is really letting the software down. We've been crying out for a machine that can deal with the processing that uncompressed video and HD requires. And, of course, a large proportion of that requirement is in a RAID array. Apple would be stupid indeed not to help drive sales of their XRAID in this way for video pros.

I'm excited - it really does make sense.

MOFS

May 31, 2003, 08:00 AM

NB: Apologies to anyone else who's written anything like this.

Whilst reading this article and the one before it mentioning the other ports, and alot of people thinking that this motherboard is for the Xserve/ Xstation, something struck me - what is the point of this processor hae built in 5.1 surround sound and USB 1.1 if its going to run as a server? I may be only a university student, but don't servers just sit in a corner and don't do actual work (not counting keeping the network running), so wouldn't the 5.1 surround sound be superfluous?:confused: :(

Postal

May 31, 2003, 08:43 AM

Here's a question: is there actually a Fibre Channel chipset on the mainboard, or are they only seeing connectors? It's entirely possible that the mainboard has a pass-through for when a Fibre Channel card is actually installed.

MB never specifically said that there was an integrated Fibre Channel chipset, so there could be a daughter card. And yes, this could be a specific PowerMac/Xserve board for the people who choose Fibre Channel, rather than the one that everybody gets. It could even be just a test board that includes Fibre Channel as a matter of course.

AidenShaw

May 31, 2003, 09:50 AM

Originally posted by Postal
Here's a question: is there actually a Fibre Channel chipset on the mainboard, or are they only seeing connectors?

I find it suspicious that the form factor of the "mystery" connectors isn't described at all.

Are they RCA jacks? Surely the MacB folks would recognize an RCA jack, and would have told us that there were "3+2" RCA jacks. (Makes component video+audio a good possibility.)

Are they mini-jacks (2.5mm headphone type)? Are they round? Are they square or rectangular.

Apple is using copper "Fibre Channel" in the Xserve RAID, HSSDC2 on the disk unit and SFP on the PCI card. These would be rectangular jacks, slightly larger than a USB port but without the off-center key that the USB ports use.

The statement that "one or two" are FC is strange - you have a pretty unique connector, and you should be able to say whether you have two matching connectors or a single one of that style.

Maybe if they were working from drawings or fuzzy photos this makes sense, but find it difficult to believe that they've spent any time with the actual board.

I'm leaning towards the theory that MacB is making all this up just to stir up a frenzy of page hits. I can't wait to see the stories that show up in the next few weeks leading up to 23 June!

alandail

May 31, 2003, 10:06 AM

Originally posted by centauratlas
>Why would you want to have a RAID system hooked up to a desktop <

And 20 years from now we'll be saying "who thought 63-bit addressing would be enough for everything? We need a 128 bit processor to handle more memory."

while I agree with pretty much everything else you said, I think it'll be much, much longer than 20 years before the above is true. 64 bits didn't just double address space, it squared it - 64 bits can address 4 billion times the 4 gigabytes that 32 bits addresses, or 16 billion gigabytes of virtual memory per process. That's a lot more dramatic than going from 640k to two gigabytes.

mactastic

May 31, 2003, 10:46 AM

If this is true (and thats a pretty big if in my mind still) it is for very high end machines, as the ability to have a software based RAID drive setup is already built into OS X. The FC PCI card is like $500 bucks! Who wants to add that to the cost of their computer if they won't be using it? If it was a matter of less than $100 I wouldn't mind so much, but this seems like an awful lot of bandwidth to put in a computer that may end up connecting to the internet over a dial-up modem if this is the new power mac.

Jeff Harrell

May 31, 2003, 10:57 AM

Originally posted by mcl
Sun's one of the the longest-running platforms using FC-AL, and their drivers STILL spontaneously LIP, and don't interop well with non-Sun GBICs.

This is off-topic, I guess, but I want you to know that your experience is not typical. In years past I managed labs full of SGI gear with terabytes of FC storage, and the only time we experienced unexpected LIPs were when there was actually a hardware failure happening.

And as for GBIC's... heck, they're completely interchangeable. They're just electro-optical transceivers.

AidenShaw

May 31, 2003, 11:02 AM

Originally posted by Jeff Harrell
And as for GBIC's... heck, they're completely interchangeable. They're just electro-optical transceivers.

In that case, could I send you a box of Agilent GBICs in trade for QLogic or IBM GBICs? (about 20 of them)

I've pulled all the Agilent ones out of my equipment and replaced them - too many weird errors with the Agilent transceivers.

Jeff Harrell

May 31, 2003, 11:19 AM

Originally posted by alandail
while I agree with pretty much everything else you said, I think it'll be much, much longer than 20 years before the above is true. 64 bits didn't just double address space, it squared it - 64 bits can address 4 billion times the 4 gigabytes that 32 bits addresses, or 16 billion gigabytes of virtual memory per process. That's a lot more dramatic than going from 640k to two gigabytes. Perspective time. 640 KB squared is 400 GB. Now, 400 GB is a lot of RAM, but I've seen many computer systems with more. (Not desktops, obviously. Big SGI Origins.)

So squaring the amount of necessary RAM in thirty years is just about right. At least so far. Now, whether that will still hold true in another 30 years is another question. Hell, we're already simulating some computational chemistry and bioinformatics jobs at the atomic level. How much more data can we stuff into those simulations?

We'll all be using more than 32 bits of RAM sometime between now and 2030. The question is, how much more?

(About that same time, we'll have to deal with the Y2K+38 problem anyway.)

Jeff Harrell

May 31, 2003, 11:20 AM

Originally posted by AidenShaw
In that case, could I send you a box of Agilent GBICs in trade for QLogic or IBM GBICs? (about 20 of them)

I've pulled all the Agilent ones out of my equipment and replaced them - too many weird errors with the Agilent transceivers. Sure. I've never noticed a problem with any GBIC. Sometimes they fail, sure, and sometimes other problems occur. But as for "weird errors?" Never.

Jeff Harrell

May 31, 2003, 11:27 AM

Originally posted by SunRiser
Now the rumors parts: the PPC 970 should make look the Dual 1.4 like a calculator and now the motherboard should have a FB connector in it... See the link? ;)

Um... no. There zero reason to use a hardware RAID card in a Power Mac, ATTO or otherwise. If you're doing video editing, you should be using software RAID-0. Fibre channel disk enclosures with FC-ATA bridges only are not exactly cheap, but they're widely available. If you want to spend more and buy more reliability, you can use native FC drives.

So: Power Mac, PCI FC card, external FC-ATA or FC-FC array. Software disk striping through Disk Utility. Poof. Done.

In a perfect world, you'd have efficient software RAID-3, but Disk Utility currently doesn't support that. If you want to rah-rah Apple for something, cheer for software RAID-3, not for hardware FCAL support.

bretm

May 31, 2003, 11:29 AM

Having a RAID doesn't mean your material is backed up. That is only if you stripe your setup as mirrored.

And if you want a raid, slap 2 extra atas in your mac, format them as a raid (mirrored if you like) and go to town. Much better performance for video editing, etc.

Originally posted by MarksEvilTwin
Whats so weird or useless about having a RAID system at home? Drives can fail spontaneously anytime, and yeah, if you keep good backups thats great, but we cant always be ready. My dad's HD randomly died last fall and he lost all these photos he took over vacation about a month earlier, because he hadn't had the chance to back it up at all, also some work documents. He now has two drives hooked up in a RAID so he's got these two drives mirroring eachother.

I think RAIDs are a great idea, and i dont think they should ONLY be used in pro environments. If you are willing to shell out the little bit extra for the second drive, its definitely worth it.

Mark

Jeff Harrell

May 31, 2003, 12:18 PM

Originally posted by bretm
Having a RAID doesn't mean your material is backed up. That is only if you stripe your setup as mirrored.

And if you want a raid, slap 2 extra atas in your mac, format them as a raid (mirrored if you like) and go to town. Much better performance for video editing, etc. A clarification.

I know I'm probably being an anal-retentive nit-picker, but it doesn't make any sense to say "stripe your setup as mirrored." Striping and mirroring are two different things. Also, if you mirror two drives, your performance will not go up, and will almost certainly go down slightly.

Striping data across disks means each disk gets a bit of the data. Imagine a file split up into four chunks, A-D, and written on to four disks, 1-4. Since all four writes can happen simultaneously (more or less), you get four times the performance as you would if you wrote the file in one chunk to one disk. That's striping. Of course, if any disk fails, you lose all your data, because B, C, and D are useless without A.

Mirroring, on the other hand, is writing the same file or chunk-of-file to more than one disk at the same time. That way if one disk fails, you can still get to your data by reading the other disk. However, since you're writing the same data twice (or more often), your computer has to do more work. How much more depends on how your disks are arranged. If you're mirroring across disk controllers and busses as well as drives, then you probably won't notice a performance dip when you mirror. But if you're mirroring across two drives on the same controller and bus (two daisy-chained firewires, for example), be prepared to deal with a performance hit.

Striping and mirroring are often used together. You can stripe across mirrors or you can mirror stripes. They're both pretty much the same; the difference comes up when it's time to rebuild a mirror set. Stripe-unit mirror sets rebuild faster.

So: striping = better performance, but lowered reliability. Mirroring = the same (or slightly lowered) performance, but better reliability. A combination of the two can give you both good performance and good reliability, but you need a minimum of four drives for that.

There are ways to improve the reliability of a stripe set, by using parity data calculations to increase the number of disks by one (rather than by doubling them), but those aren't supported by Disk Utility at this time. When you hear about RAID-3 and RAID-5, that's what they're talking about. (There's a RAID-4, but nobody uses it.)

Just a clarification.

rjwill246

May 31, 2003, 01:21 PM

Originally posted by bahram
"Uh folks, this is all from the site that claimed to have benchmarks of the 970 using a soon to be released new version of Bryce that is far into development. Guess what? Corel just announced that Bryce version 5 is it as far as the mac goes. Everything on the MB site is just totally and completely bogus, and specifically, news about Bryce is the final nail in the coffin in regards to their earlier claimed benchmarks"

Here is the quote from Macbidouile:

"J'aurais du vous indiquer, les benchmarks de bryce ... mais sur un beta de la version suivante de bryce, qui doit arriver en juillet ou août de cette année. La version 6 supportera des configurations multi-processeurs....La version beta de Bryce 6 est disponible sur quelques sites P2P, notablement hotline. Le vrai fichier est Corel Bryce 6 Beta.sit, et il pese a 91.2 MO. "

It clearly indicates that Bryce 6 was being tested as a beta. The type and size of file were given and there is no question that a beta version could have been being tested and that after some reflection, Corel decided to pull the plug on it. I don't think this invalidates MB's rumours carte blanche at all. Lionel, from MB is both sincere and adamant that the 970 will be sooner than later. If you have followed their announcements you will know that he is clearly hoping it will be announced in June, possibly coming out in a limited form anytime August through September and the inference is that it will appear across the line of Apple product by next year. I happen to think he is close to the mark.

Cappy

May 31, 2003, 01:28 PM

Originally posted by mactastic
Who wants to add that to the cost of their computer if they won't be using it?

This is something Apple has done for awhile. How about gigabit ethernet a few years ago when it was much more expensive? What about dvd-r burners when the purchaser doesn't want it on a highend system?

Apple does this to add more perceived value to specific price points and puts the marketing spin on it. It's nothing new.

Cappy

May 31, 2003, 01:33 PM

I only saw one person mention this in the thread and it looked like everyone just blew past it and that is that it's likely fibre for the latest incarnation of firewire. It all makes perfect sense. The specs for it point this all out and Apple is going to be one vendor who will push this as it fits perfectly into their highend line of systems.

Remember firewire really isn't about the physical media but the protocol.

ZeeOwl

May 31, 2003, 01:42 PM

Uh folks, this is all from the site that claimed to have benchmarks of the 970 using a soon to be released new version of Bryce that is far into development. Guess what? Corel just announced that Bryce version 5 is it as far as the mac goes. Everything on the MB site is just totally and completely bogus, and specifically, news about Bryce is the final nail in the coffin in regards to their earlier claimed benchmarks.

Yep. I saw the posts that say that Corel is discontinuing development of Bryce for the Mac too. Now, I'm not saying that the MacBidouille rumours are fact. But I do think they could be true anyways. The key words here are "discontinuing developement". Corel says it will continue providing support for Bryce 5 Mac, and will not release any new versions of Bryce for the Mac. It does not say that they never started working on Bryce 6 for Mac. The MacBidouille benchmarks article talks about a beta of Bryce 6.

Companies do occasionally abandon a projet under development. To those familiar with Canadian history, remember the Avro Arrow (CF-105)? To those that aren't, it was a twin-engine jet interceptor built by Avro Canada in the late 50's for the Canadian Air Force, with a top speed of Mach 2.5. Yup, you read right, Mach 2.5 in the late 50's. 5 working prototypes were built. Just before actual production started, the project was scrapped, and the prototypes and all related documentation were destroyed.

mcl

May 31, 2003, 01:46 PM

Originally posted by illumin8
Bzzt. Thanks for playing. You obviously don't know jack ***** about fibre channel. All of the major Unix players use fibre channel for high capacity storage, and I assure you it works quite well. This includes Sun, HP, IBM, Fujitsu, SGI, and anyone else worth a damn.

Yeah, okay. I don't know anything about it. The 10 years I've been working with it professionally are all just a hallucination. The production beta-testing I did on EMC's FC-AL driver for Solaris back in '96 was a hallucination. The fact that I use fibre channel professionally on a daily basis is just a hallucination.

By the way, you might want to brush up on your reading and comprehension skills. I never said that FC wasn't in common use. I said it was still buggy. And for external FC-AL (not the internal FC-AL which is common on high end workstation and server gear, and which is not germaine to this conversation, because this conversation is dealing with EXTERNAL FC-AL connectors), it still is.
But hell, don't believe me. Go read through SunSolve's bug database for the FC-AL drivers. You *do* have a SunSolve account, don't you?

By the way, fibre channel now supports 2gb a second transfer rate, and if by "short-haul" you mean several hundred meters then I guess that would be short.

Yes, that fits my definition of "short-haul". My definition of "long-haul" spans oceans and continents.

Get with the program. Fibre channel is superior to all flavors of SCSI, except perhaps SCSI-320, in speed, and the ability to have up to 127 devices on a single fibre channel loop is a plus.

that's really, really, amazing. I completely slept through the part where I went off on a tirade about the performance and feature-set of FC, necessitating your rah-rah-ing of both.

If you insist on arguing (particularly when you're not in posession of facts that would justify a personal attack on my knowledge and background ), learn how to argue effectively. Here's a free basic skill: Stick to the facts, avoid logical fallacies, and address only those points your opponent has raised.

You fail on all three criteria.

HTH. HAND.

mcl

May 31, 2003, 01:54 PM

Originally posted by Jeff Harrell
This is off-topic, I guess, but I want you to know that your experience is not typical. In years past I managed labs full of SGI gear with terabytes of FC storage, and the only time we experienced unexpected LIPs were when there was actually a hardware failure happening.

And as for GBIC's... heck, they're completely interchangeable. They're just electro-optical transceivers.

My experience as described in the bit you quote was with Sun gear. How, exactly, does an argument from SGI gear make that expereince atypical? In years past I managed the systems for the nation's largest nuclear power company, with terabytes of FC storage, and we experienced unexpected LIPs quite regularly, particularly when the server was a Sun, and the storage was non-Sun (typically, but not always, EMC).

And as for the GBICs, they're interghangeable because there's a hardware design standard. As I'm sure you're well aware, individual manufacturers do not hew strictly to said standard. Some systems are more sensitive than others to variations from the standard. Suns are, again, notorious for this. Sun GBICs work fine. Non-Sun GBICs are a crapshoot. And certain vendors (e.g., EMC) have a history of requiring that their equipment use their GBICs on both ends of the loop before the system will qualify for support (In fairness, I do believe EMC has stopped this nonsense, but it used to be the case).

ZeeOwl

May 31, 2003, 01:54 PM

Just as an FYI, this is a human translation of the MacBidouille article by your's truly. I speak French fluently, so the wording is an exact English equivalent as far as meaning is concerned.

"We have received information on some of the mysterious external connectors on the PPC 970 motherboard. One of them (and probably 2) is a Fibre-channel connector to connect, for example, an XServe RAID."

mcl

May 31, 2003, 01:57 PM

Originally posted by Jeff Harrell
A clarification.
There are ways to improve the reliability of a stripe set, by using parity data calculations to increase the number of disks by one (rather than by doubling them), but those aren't supported by Disk Utility at this time. When you hear about RAID-3 and RAID-5, that's what they're talking about. (There's a RAID-4, but nobody uses it.)

Just a clarification.

As long as we're nitpicking, it's RAID 3 that nobody uses. RAID 4 is in common use, and you need look no further than NetApp's filesystem to find it.

mcl

May 31, 2003, 02:03 PM

Originally posted by bretm
Having a RAID doesn't mean your material is backed up. That is only if you stripe your setup as mirrored.

having a mirror doesn't mean your material is backed up. I've seen mirrors fail, bad data from one set get copied to the good mirror, multiple simultaneous disk failures, etc.

Having level 0 backups on tape stored in multiple geographically-diverse locations and checked for readability and integrity on a regular basis means your material is backed up.

Hainvg a mirror just means you're hoping a disk or three won't fail.

mcl

May 31, 2003, 02:07 PM

Originally posted by Jeff Harrell

In a perfect world, you'd have efficient software RAID-3, but Disk Utility currently doesn't support that. If you want to rah-rah Apple for something, cheer for software RAID-3, not for hardware FCAL support.

How, exactly, is software-based RAID preferable to hardware-based RAID? It's certainly not more efficient, and I would hesitate to apply the term "efficient" to software-based RAID.

As for RAID 3, see my previous post on the matter. It's not in use. Even if we were discussing RAID 4 (the version of that approach that IS in use), you'd certainly want hardware-based RAID 4, not software-based. Software-based RAID 4 would introduce unnecessary overhead on every write, and would place an unnecessary burden on the system's CPU.

Hardware-based RAID is in widespread use. Why cripple a system, performance-wise, by doing in system memory what's commonly and relatively cheaply done in firmware?

ZeeOwl

May 31, 2003, 02:31 PM

This is just IMHO. I think built-in Fibre-channel makes perfect sense, coming from Steve Jobs/Apple. The PowerMac is not marketed as as a home desktop computer. The eMac and flat-panel iMac are. The PowerMac is meant to be a pro (and might I add, high-end specialized multimedia capable) computer. Remember when Apple first started shipping PowerMacs with built-in FireWire? How much was as add-on FireWire PCI-card back then? Several hundred dollars... How common and affordable were FireWire equipped peripherals? All there was were high-end, expensive miniDV and pro DV cameras. This was how many years ago? Maybe 4? And now... Hey, you can even buy an MP3 player with FireWire! ;)

People, stop thinking like the one's who created the Y2K fiasco. And stop expecting Apple to "make perfect marketing sense". Apple innovates, and the rest of the world follows. Not the other way around.

I think the person who mentioned FireWire over Fibre-channel is on to something. This would play right into Apple's style. They did develop FireWire, after all. And if they could manage a 3.2 Gb/s network connection, that would be a phenomenal marketing advantage. This might also be an explanation for no mention of an Ethernet port. Maybe they've done away with it. Want 10/100/1000 Gb/s Ehernet? Plug an adapter into one of the Fibre-channel ports. Like they replaced VGA with ADC. Hook up two PowerMac 970's directly, and you've got a 3.2 Gb/s network. What more could you want in a video/audio workstation? And I was thinking (fantasizing?), if this thing has two Fibre-channel ports (as MacBidouille suggests), for a small network you wouldn't even need a hub. Just daisy-chain the machines together. Cool. :D

AidenShaw

May 31, 2003, 02:33 PM

Originally posted by mcl
As for RAID 3, see my previous post on the matter. It's not in use. Even if we were discussing RAID 4 (the version of that approach that IS in use

Isn't RAID-5 the most common parity-based RAID? My hardware RAID boxes do 0, 1 and 5 (plus combinations like 10, 01, 50, ...).

Google:

RAID-3 - 21,500 hits
RAID-4 - 17,700 hits
RAID-5 - 155,000 hits

Look at http://www.jadlog.demon.co.uk/raid3.htm - RAID-4 not even mentioned.

andyduncan

May 31, 2003, 02:38 PM

Originally posted by ZeeOwl
Companies do occasionally abandon a projet under development.

No they don't. My Power Express, my Orient Express, my Hyperbolic and my StarMax 6000 are all on their way. really.

dguisinger

May 31, 2003, 02:50 PM

I still think its possible. No one has yet countered my opinion that building it into their chipsets would reduce the cost significantly from using the $500 add-on boards. This is what they did with Gigabit Ethernet as well.

Infact, I would not be surprised if Apple's new southbridge would include:

- Surround Sound
- Dual Gig Ethernet
- ATA-133 (Serial ATA is rumored to be a revision off)
- USB 2
- Fibre Channel

There are many PC motherboards (from Tyan for example) for high-end applications that already include dual gig ethernet, usb2 and S-ATA RAID along with dual Opteron support. I do not see it too hard for apple to develop the chip above.

Infact, developing it allows apple to bring the features to as many machines as quickly as possible, while reducing overhead. If they add the features to the xserve, they might as well put them in the PowerMac. It costs way too much for a company of Apple's size to develop multiple southbridge chips to be in use simultaniously.

mcl

May 31, 2003, 02:57 PM

Originally posted by AidenShaw
Isn't RAID-5 the most common parity-based RAID? My hardware RAID boxes do 0, 1 and 5 (plus combinations like 10, 01, 50, ...).

Yes, it is. However, we were discussing (the absence of) RAID 3.

Look at http://www.jadlog.demon.co.uk/raid3.htm - RAID-4 not even mentioned.

Yes. That's because the RAID 3 they describe IS RAID 4. RAID 3 has a different approach to parity writes.

Perhaps the industry started referring to RAID 4 as RAID 3 when I wasn't looking, and stopped mentioning RAID 4 altogether. However, I assure you that RAID 4 (as distinguished from RAID 3) is what's in use today when people claim they're using RAID 3. Again, one need look no further than NetApp's filesystem.
And that information comes from one of the people who wrote WAFL, and was backed up by quite a bit of evidence.

RAID 3 simply wouldn't ever actually be used. The performance sucks, because the drives all do synchronous writes. It kills transaction-based I/O, and is therefore fairly worthless for real-world use.

If you doubt the existence of RAID 4, I refer you to "A Case for Redundant Arrays of Inexpensive Disks (RAID)" Patterson, Gibson, & Katz (1988), presented at the 1988 conference of the ACM Special Interest Group on Management of Data (SIGMOD). That's the paper that defined RAID.

And, once again, I'm not saying anything about RAID 5. It exists, I've used it for years, it's perfectly suited to certain tasks, and unsuitable for others. However, the topic of conversation is RAID 3/4, and which one is in use today, not 5. Nobody's questioning the existence or widespread use of RAID 5.

Jeff Harrell

May 31, 2003, 03:15 PM

Originally posted by mcl
How, exactly, is software-based RAID preferable to hardware-based RAID?If you're more interested in bandwidth than IOPS, the difference between hardware and software RAID is not significant. Compare a hardware RAID from, say, HDS to an XVM on IRIX, both with RAID 0+1. The XVM is faster. I've done the tests myself.

As for RAID 3, see my previous post on the matter. It's not in use.Uh, sorry, dude. You're just plain wrong about this one. Stone+Wire is software RAID-3. Arguably there's no more high-performance sequential read-write filesystem in the world.

Hardware-based RAID is in widespread use. Why cripple a system, performance-wise, by doing in system memory what's commonly and relatively cheaply done in firmware? Because it's neither more common nor cheaper to do it with a hardware controller. I mean, if you want to use an incredibly crappy controller, sure, but if you're comparing a reliable system to another reliable system, think software RAID-0+1 or RAID-3 to hardware RAID 0+1 or RAID-3. For servers, sure, think hardware storage management. For desktop (or at least single-user) use, think software RAID. Dollar per unit value, that's how it works out.

Jeff Harrell

May 31, 2003, 03:17 PM

Originally posted by AidenShaw
Isn't RAID-5 the most common parity-based RAID?

In my experience, RAID-3 is more common. It depends on whether you're doing sequential read/write or sparse. If you're talking about a file or database server, RAID-5 is a good choice. If you're talking about video playback or editing, RAID-3 wins.

Jeff Harrell

May 31, 2003, 03:20 PM

Originally posted by mcl
Yes. That's because the RAID 3 they describe IS RAID 4. RAID 3 has a different approach to parity writes.Um. No. RAID 4 and 5 distribute parity data across all drives in a stripe set. Four and five differ in the units they use for calculating parity. RAID 3 on the other hand writes data to (for example) 1, 2, 3, and 4, and parity to 5. RAID 4/5 uses striped parity, while RAID 3 uses a dedicated parity drive. The differences can be really confusing.

mcl

May 31, 2003, 03:23 PM

Originally posted by Jeff Harrell
In my experience, RAID-3 is more common. It depends on whether you're doing sequential read/write or sparse. If you're talking about a file or database server, RAID-5 is a good choice. If you're talking about video playback or editing, RAID-3 wins.

RAID 5 for databases?!? Never! The performance hits on writes make it ridiculously slow for high-TPS DBs!

If you're going to run a DB with RAID storage, you should go 0+1.

mcl

May 31, 2003, 03:25 PM

Originally posted by Jeff Harrell
Um. No. RAID 4 and 5 distribute parity data across all drives in a stripe set. Four and five differ in the units they use for calculating parity. RAID 3 on the other hand writes data to (for example) 1, 2, 3, and 4, and parity to 5. RAID 4/5 uses striped parity, while RAID 3 uses a dedicated parity drive. The differences can be really confusing.

Look, you're simply wrong. If you want to debate what RAID 3,4, and 5 do, go read the paper I cited earlier. It's the authoritative source. Then come back and tell me RAID 4 doesn't use a dedicated parity disk.

The differences are only confusing when you mistake marketing pap for facts. Stop listening to the people in the ties, and start listening to the people that write the code, build the boards, and run the systems.

ZeeOwl

May 31, 2003, 03:37 PM

Originally posted by andyduncan
No they don't. My Power Express, my Orient Express, my Hyperbolic and my StarMax 6000 are all on their way. really.

lol

mcl

May 31, 2003, 03:39 PM

Originally posted by Jeff Harrell
If you're more interested in bandwidth than IOPS, the difference between hardware and software RAID is not significant. Compare a hardware RAID from, say, HDS to an XVM on IRIX, both with RAID 0+1. The XVM is faster. I've done the tests myself.

Differences in hardware vs. software RAID for 0, 1, 1+0 and 0+1 are negligible. Your results are entirely unsurprising. I'd like to see figures supporting such a claim for RAID 4 or RAID 5, however.

I don't actually expect you to produce them, because I know you can't demonstrate faster transactions for software RAID 4 or 5 vs. hardware. The hardware-based solution will win when there's parity computations to consider. Either you know this and you're trying to be deceptive, or you don't know this and you're arguing from atop very shaky parapet.

Uh, sorry, dude. You're just plain wrong about this one. Stone+Wire is software RAID-3. Arguably there's no more high-performance sequential read-write filesystem in the world.

Really? Then why do they also offer hardware RAID 5 in the same platform, and tout it as higher-performance than their so-called "RAID 3"?

Since Discreet's website layout leaves something to be desired, please point me to a tech note, manual, or white paper that describes the parity writes of their RAID 3, and I'll show you the part in the Berkeley paper that classifies that particular use as RAID 4, and I'll even explain to you why it's not RAID 3. Until then, the only evidence I see supporting your claim that it's RAID 3 is that Discreet calls it RAID 3, therefore it must be true.

Oh, and why don't I see Stone+Wire RAID 3 storage attached to the mission-critical servers in any of the world's largest financial houses, digital movie studios, financial markets, Fortune 500 companies, etc?

If they're that good, people would be stumbling all over each other to buy and use them. They're not.

(and when I say, "why don't *I* see, I do mean me. I have had occasion to work with the folks who spec, buy, install, and maintain that equipment, or work on it directly. I have first-hand knowledge of the systems, both desktop and production server, in Pixar, Dreamworks, and ILM, as well as many of the largest, most well-known high-tech companies in the Valley, and several out East and in the Midwest.)

Oh, and I'm not your "dude". Is faux-familiarity somehow supposed to bolster your argument and distract from a glaring lack of supporting evidence?

For servers, sure, think hardware storage management. For desktop (or at least single-user) use, think software RAID. Dollar per unit value, that's how it works out.

Yes, I'll agree that it's more cost-effective to use software-based RAID on the desktop. However, neither you nor I were arguing from that position. We were both discussing enterprise-class solutions. Whipping out software-based RAID on the desktop as justification for your position at this late point in the discussion is disingenuous at best.

andyduncan

May 31, 2003, 03:57 PM

Originally posted by mcl
Yes. That's because the RAID 3 they describe IS RAID 4. RAID 3 has a different approach to parity writes.

Actually I think they are correctly describing both RAID 3 and RAID 4. The difference between the two being the way data is striped, either at the block level or the byte level. This fact is described elsewhere, including the paper you reference:
Page 113, regarding RAID 4's changes over RAID 3: "We no longer spread the individual transfer information across several disks, but keep each individual unit in a single disk."

The parity is calculated differently, but the site doesn't get into the algorithm, and the layout of disks they show could be either 3 or 4.

The only thing that would imply they are actually talking about RAID 3 rather than RAID 4 is that they recommend this type of RAID for Video/image editing etc. RAID 4's advantage over RAID 3 is small file read access, a quality that would not benefit those applications.

edit: grammar.

mcl

May 31, 2003, 04:07 PM

Originally posted by andyduncan
Actually I think they are correctly describing both RAID 3 and RAID 4. The difference between the two being the way data is striped, either at the block level or the byte level. This fact is described elsewhere, including the paper you reference:
Page 113, regarding RAID 4's changes over RAID 3: "We no longer spread the individual transfer information across several disks, but keep each individual unit in a single disk."

The parity is calculated differently, but the site doesn't get into the algorithm, and the layout of disks they show could be either 3 or 4.

The only thing that would imply they are actually talking about RAID 3 rather than RAID 4 is that they recommend this type of RAID for Video/image editing etc. RAID 4's advantage over RAID 3 is small file read access, a quality that would not benefit those applications.

edit: grammar.

Untrue. RAID 3 also suffers in comparison to RAID 4 on writes, because RAID 3 writes are all disk-synchronous, so writes (always a short-wait event in the kernel, and thus problematic if delayed) are delayed until the data for all disks is ready to be written. Thus, each RAID 3 write takes more time than each equivalent RAID 4 write. In almost every case, substantially more time. RAID 3 is the worst-performing of all the RAID levels in terms of writes, and this is why it is not in use today.

As for the parity computations, RAID 3 computes bytewise parity across all disks, whereas the parity on RAID 4 is a delta of blockwise parity -- a much more efficient algorithm.

Having said all of this, and realizing that most of you are arguing from the point-of-view of video editing, whereas I'm arguing from a position of general enterprise-class mission-critical application, I think it's important to point out that both RAID 3 *AND* RAID 4 suffer from bottlenecking at the spindle of the parity disk. Interleaved parity, such as that found in RAID 5, offers better performance than either for reads. Issues regarding filesize can be addressed easily by increasing the default block read and write size on the filesystem written on top of the RAID layout. I wouldn't be surprised if several of you have been evaluating disk performance without ever changing the default filesystem parameters (and where possible, physical data layout on the platter, a la AIX) to more closely match the application to which the disks will be put.

I'm tired of arguing. The vendors have obviously been marketing "RAID 3" to A/V production houses, for whatever reason. I can't argue with marketing...it is what it is. Further, without concrete, detailed information on the striping and parity computations involved, such arguments quickly become futile, because it's impossible to debate without that necessary common ground.

If someone's got some hard, real-world (not based on the 1990 Berkeley simulator, and not ripped from a RAID vendor's marketing materials) data demonstrating RAID 3 (synchronous parallel reads/writes with bytewise parity on a dedicated disk) performance on small and large block reads and writes, for software and hardware RAID implementations, outperforming RAID 4 and/or RAID 5, I welcome you to introduce it.

If all you've got is a paper touting the benefits of RAID 3, without any meaningful comparison to the other RAID levels, or without the software/hardware comparison, or without the comparison on small-block v. large-block reads/writes, don't bother.

void

May 31, 2003, 04:26 PM

[mod edit]don't even start[/mod edit]
Hey man! It was all in good fun! Damn you Duke!:D

Dunepilot

May 31, 2003, 04:45 PM

off-topic:

There is clearly an abundance of intelligent posters on the boards for once.

I don't know anything about RAID at all, but I'd just like to ask everyone to try and be a little less heated - this all feels a little to bitter at the moment.

Peace to all (I don't often feel like this):D

macosr

May 31, 2003, 06:24 PM

If this is FC there are VERY few reasons to use this over SCSI 160 or 320 unless you are running a SAN cluster in Active/Active/...

There really isn't that much more throughput to an individual machine when using FC for storage.

Now if Apple is just bringing FC to the masses as they did GbEnet to bring prices down then great! However, I don't see too many consumers spending ~$30,000 on a switched FC storage network. Also, if there is only 1 port and 1 channel then it isn't very useful anyhow as nearly everyone wants redundancy (multiple switched) when putting FC into place.

Also, the comments regarding software RAID being as fast as hardware RAID is just plain wrong. Try and get 600Mb/sec transfer with software RAID on a heavily loaded machine.

Jeff Harrell

May 31, 2003, 06:49 PM

Originally posted by macosr
If this is FC there are VERY few reasons to use this over SCSI 160 or 320 unless you are running a SAN cluster in Active/Active/...

More than 100 devices per loop. Cheap optical cables. Multi-kilometer cable runs. Switches for storage consolidation. There are lots of reasons why Fibre Channel is neat other than performance. It may or may not be what you need, but it's not a one-to-one comparison with SCSI.

Now if Apple is just bringing FC to the masses as they did GbEnet to bring prices down then great!

Problem is, Apple's thrown its weight behind FireWire, which pretty much is a one-to-one comparison with FCAL. I know there's FireWire 3200 gear out there, although I don't know if it's prototype-stage or incredibly-expensive-production-model stage. Both run over fiber optic cables (cheap and far!), both support switches and hubs for managed topologies, both are about the same speed, give or take a gigabit per second. Why would Apple built FCAL into the machine when they'd prefer FireWire rise to challenge FCAL?

Try and get 600Mb/sec transfer with software RAID on a heavily loaded machine.

Both been there and done that. Maybe you're comparing hardware RAID to crappy software RAID. XVM on IRIX is the bee's knees. I've personally seen a filesystem write (write!) more than 2 GB/s. That's bytes with a capital B. Really big system, lots of disks. Not a hardware controller in sight. The configuration was RAID 0+1. In my own lab, smaller machines were comparable, doing more than 95 MB/s reads and writes off of eight FC disks (Seagates, if I remember correctly) through XVM. You can't get much faster on a single FCAL loop.

Is Apple's striping/mirroring implementation anywhere near that good? Don't know. But it's not safe to naturally assume it isn't, either.

dguisinger

May 31, 2003, 07:05 PM

Originally posted by Jeff Harrell
More than 100 devices per loop. Cheap optical cables. Multi-kilometer cable runs. Switches for storage consolidation. There are lots of reasons why Fibre Channel is neat other than performance. It may or may not be what you need, but it's not a one-to-one comparison with SCSI.

Problem is, Apple's thrown its weight behind FireWire, which pretty much is a one-to-one comparison with FCAL. I know there's FireWire 3200 gear out there, although I don't know if it's prototype-stage or incredibly-expensive-production-model stage. Both run over fiber optic cables (cheap and far!), both support switches and hubs for managed topologies, both are about the same speed, give or take a gigabit per second. Why would Apple built FCAL into the machine when they'd prefer FireWire rise to challenge FCAL?

Both been there and done that. Maybe you're comparing hardware RAID to crappy software RAID. XVM on IRIX is the bee's knees. I've personally seen a filesystem write (write!) more than 2 GB/s. That's bytes with a capital B. Really big system, lots of disks. Not a hardware controller in sight. The configuration was RAID 0+1. In my own lab, smaller machines were comparable, doing more than 95 MB/s reads and writes off of eight FC disks (Seagates, if I remember correctly) through XVM. You can't get much faster on a single FCAL loop.

Is Apple's striping/mirroring implementation anywhere near that good? Don't know. But it's not safe to naturally assume it isn't, either.

FireWire is not a direct competitor to FibreChannel. FibreChannel is designed for storage networks only. FireWire is more general purpose, and does not yet push such high speeds. Even FW800 doesn't come near in disk performance, especially when you go with dual channel 2Gbp FibreChannel.

fiardinkum

Jun 1, 2003, 07:26 AM

Assuming this is the motherboard for the new PowerMac (or whatever it will be called) this connector is more likely a TosLink optical connector, allowing output of digital sound and DTS/5.1 to be sent to an Amplifier. I'm hoping on this one!! Why would Apple build fibrechannel into a motherboard when only a few people would use it and when PCI cards are readily available to do the same thing?

macosr

Jun 1, 2003, 04:49 PM

Originally posted by Jeff Harrell
More than 100 devices per loop. Cheap optical cables. Multi-kilometer cable runs. Switches for storage consolidation. There are lots of reasons why Fibre Channel is neat other than performance. It may or may not be what you need, but it's not a one-to-one comparison with SCSI.

Problem is, Apple's thrown its weight behind FireWire, which pretty much is a one-to-one comparison with FCAL. I know there's FireWire 3200 gear out there, although I don't know if it's prototype-stage or incredibly-expensive-production-model stage. Both run over fiber optic cables (cheap and far!), both support switches and hubs for managed topologies, both are about the same speed, give or take a gigabit per second. Why would Apple built FCAL into the machine when they'd prefer FireWire rise to challenge FCAL?

Both been there and done that. Maybe you're comparing hardware RAID to crappy software RAID. XVM on IRIX is the bee's knees. I've personally seen a filesystem write (write!) more than 2 GB/s. That's bytes with a capital B. Really big system, lots of disks. Not a hardware controller in sight. The configuration was RAID 0+1. In my own lab, smaller machines were comparable, doing more than 95 MB/s reads and writes off of eight FC disks (Seagates, if I remember correctly) through XVM. You can't get much faster on a single FCAL loop.

Is Apple's striping/mirroring implementation anywhere near that good? Don't know. But it's not safe to naturally assume it isn't, either.

I completely agree with your assesment on FC and the other features. Here is the real question..."will all home users that will use these features please raise your hand." NONE

Firewire != FC

Regarding RAID throughput. I won't argue it...I was just pointing out what is true. BTW, we get the ~600Mb/sec on RAID 0+1 with a total of 4 drives. If you have some secrete way of doing software RAID then maybe you should contact Dell, Sun, HP, etc so they could some some money on their hardware RAID setups :D

elvisizer

Jun 1, 2003, 09:07 PM

<rant> altogether now, people, say it- FIBER does NOT equal RAID.
or, to put it another way, you don't eed fiber to set up a raid, so will everyone posting saying that it makes sense for apple to put fiber on a powermac mb so that people can use RAID arrays please stop it! yr makin' us mac users look like dips. setting up a raid array on a mac does NOT require fiber, it just requires at least 2 identical capacity HD's. that's it. so go buy 2 drives, turn 'em into a RAID with OSX disk utility, and forget about having fiber on your next powermac mb, 'cause that just ain't gonna happen. </rant>

personally, i think this means that if this rumored mb does exist, it must be headed for a new xserve. that's the only thing that makes sense. either that or it's all hooey.

illumin8

Jun 1, 2003, 10:39 PM

Originally posted by Jeff Harrell
In a perfect world, you'd have efficient software RAID-3, but Disk Utility currently doesn't support that. If you want to rah-rah Apple for something, cheer for software RAID-3, not for hardware FCAL support. I agreed with everything you said up until that point. If you're doing video editing or anything which requires high speed write performance you don't want to be using any flavor of RAID that calculates parity... That's the problem with RAID flavors 3, 4, and 5. The parity calculations are a bitch and you take a huge performance hit on writes.

Jeff Harrell

Jun 1, 2003, 10:43 PM

Originally posted by illumin8
I agreed with everything you said up until that point. If you're doing video editing or anything which requires high speed write performance you don't want to be using any flavor of RAID that calculates parity... That's the problem with RAID flavors 3, 4, and 5. The parity calculations are a bitch and you take a huge performance hit on writes. That isn't necessarily true. (See, for example, the aforementioned Stone+Wire.) And the extra work necessary to handle parity is more than worth it when you consider the alternative. Let's say you've got 32 drives attached to your machine for storing HD video. They're striped. You lose one disk. Now you have to rebuild and repopulate the entire array. The other option is to double the number of drives to 64 and mirror them, which would cost a fortune.

Suddenly, RAID-3 sounds like a pretty good idea.

illumin8

Jun 1, 2003, 10:45 PM

Originally posted by Postal
Here's a question: is there actually a Fibre Channel chipset on the mainboard, or are they only seeing connectors? It's entirely possible that the mainboard has a pass-through for when a Fibre Channel card is actually installed.

MB never specifically said that there was an integrated Fibre Channel chipset, so there could be a daughter card. And yes, this could be a specific PowerMac/Xserve board for the people who choose Fibre Channel, rather than the one that everybody gets. It could even be just a test board that includes Fibre Channel as a matter of course. Wow, that's a very insightful post. You see, there is a definite possibility that Apple would want to include a FC-AL controller that would have higher integration with the motherboard. Just to give you an idea of the bandwidth requirements:

64-bit 33mhz. PCI slot can push 256 MBytes a second bandwidth. These are the slots currently used on the PowerMacs. The new FCAL controllers are 2gbps, or roughly 250 Mbytes per second, however, they are full duplex, so you could be pushing 500 Mbytes per second through them if you are reading and writing. This would clog your PCI bus. You could switch to 64-bit 66mhz. PCI which would double your bandwidth to 512Mbytes per second, however, I'm guessing that Apple might decide to do something better:

Chain an FC-AL controller directly off of a Southbridge on the motherboard. Give it a dedicated channel to the CPU through the chipset, just like AGP graphics. This eliminates bottlenecks on your PCI bus, and by allowing you to add a daughterboard to your existing PowerMac motherboard, doesn't increase the cost too much for low-end users that don't need to push 500+ Mbytes a second out to disk.

This is starting to make sense now. It still is some awfully powerful hardware that less than 1% of Apple's market probably needs, but for the video professionals that need it, it is essential.

illumin8

Jun 1, 2003, 10:57 PM

Originally posted by mcl
The production beta-testing I did on EMC's FC-AL driver for Solaris back in '96 was a hallucination.Perhaps that's why you perceive FC as being buggy. If you were working on beta drivers for EMC, I can imagine you might think that.
I never said that FC wasn't in common use. I said it was still buggy. And for external FC-AL (not the internal FC-AL which is common on high end workstation and server gear, and which is not germaine to this conversation, because this conversation is dealing with EXTERNAL FC-AL connectors), it still is.This is not true. Do you think banks, telcos, airlines, and other Fortune 100 companies would trust their largest databases and applications to storage that was buggy? Sure, there are occasionally issues, but I personally work on Sun servers at a data warehouse controlling 20 TB (that's Terabytes) of FCAL storage and they have not had unscheduled downtime for the last 3 years. I also work on Sun servers at a Telco using 10 TB of FCAL storage. Do you think these companies would trust their most mission-critical apps (for the telco it's their billing database, tracking every cellphone number dialed by every customer) to an unproven and "buggy" technology?
Yes, that fits my definition of "short-haul". My definition of "long-haul" spans oceans and continents.F-cking ridiculous... Who on earth wants their disks to be sitting on a different continent than their server? There is no need for such technology.

illumin8

Jun 1, 2003, 11:06 PM

Originally posted by mcl
In years past I managed the systems for the nation's largest nuclear power company, with terabytes of FC storage, and we experienced unexpected LIPs quite regularly, particularly when the server was a Sun, and the storage was non-Sun (typically, but not always, EMC).Just for the record I think it's unfair to make a comparison between today's various Fibre Channel implementations and those from years ago. As with any new technology, there have been bugs and they have been resolved. In my experience it doesn't pay to be on the bleeding edge of any technology.

I too have seen issues with using Sun servers on EMC storage, and they can usually be resolved by one of four things:

1. Updating firmward on your HBA.
2. Update the HBA's driver in Solaris.
3. Update microcode on the EMC array.
4. Patch the OS or edit your /etc/system file, for example SET_LWP_STACK_SIZE being set incorrectly by Veritas can definitely cause the system to panic.

As always, if you follow the manufacturer's recommendations you will probably be fine. Now that Sun has an enterprise storage offering I highly recommend that you check it out (obligatory plug:) . It is based on Hitachi storage and has been tested to work extremely well with our servers.

BTW, did you work for Entergy? Just curious, I know a couple IT guys that work there.

illumin8

Jun 1, 2003, 11:19 PM

Originally posted by Jeff Harrell
I've personally seen a filesystem write (write!) more than 2 GB/s. That's bytes with a capital B.

Cool! I've never seen a filesystem write that fast before. I'm just curious how you accomplished all that bandwidth out to disk?

Since the fastest FCAL controllers right now are 2Gbps (gigabits, not bytes), it would have taken what, at least 8 of them in the same server connected to the same array to accomplish this? Not to mention PCI bandwidth.

I'm not saying I don't believe you, I'm just curious to know the details, being kind of a server geek myself.

Sun Baked

Jun 1, 2003, 11:20 PM

Originally posted by illumin8
Just to give you an idea of the bandwidth requirements:

64-bit 33mhz. PCI slot can push 256 MBytes a second bandwidth. These are the slots currently used on the PowerMacs. The new FCAL controllers are 2gbps, or roughly 250 Mbytes per second, however, they are full duplex, so you could be pushing 500 Mbytes per second through them if you are reading and writing. This would clog your PCI bus. You could switch to 64-bit 66mhz. PCI which would double your bandwidth to 512Mbytes per second, however, I'm guessing that Apple might decide to do something better: Currently the X-Serves/Cluster Nodes are running...

X-Serves - Two full-length 64-bit, 66MHz PCI slots and one half-length 32-bit, 66Mhz combination PCI/AGP slot

Cluster Node - Two 64-bit, 66MHz PCI slots

illumin8

Jun 1, 2003, 11:25 PM

Originally posted by Sun Baked
Currently the X-Serves/Cluster Nodes are running...

X-Serves - Two full-length 64-bit, 66MHz PCI slots and one half-length 32-bit, 66Mhz combination PCI/AGP slot

Cluster Node - Two 64-bit, 66MHz PCI slots Do you know how many PCI busses they have? If both 64-bit slots are tied to the same PCI bus then they are sharing that 512Mbytes a second bandwidth.

My point is that Apple seems to be tying more and more system critical functions into the Southbridge. Audio, Gigabit Ethernet, USB, Firewire, and now FC-AL. This would make sense because then the most common peripherals have access to a lot more bandwidth, and your Avid or Protools hardware can have the PCI bandwidth all to itself.

Jeff Harrell

Jun 1, 2003, 11:25 PM

Originally posted by illumin8
Cool! I've never seen a filesystem write that fast before. I'm just curious how you accomplished all that bandwidth out to disk?

Since the fastest FCAL controllers right now are 2Gbps (gigabits, not bytes), it would have taken what, at least 8 of them in the same server connected to the same array to accomplish this? Not to mention PCI bandwidth.

I'm not saying I don't believe you, I'm just curious to know the details, being kind of a server geek myself. I don't remember the precise configuration. It was an Origin 2000 with 64 CPU's. It think it was 64 CPU's; it was four racks, so it could have been as few as 16 or as many as 64. But I think it was fully populated.

The disk was Clariion FC RAID. I forget the precise model; it was the one SGI OEM'd, so we just called it FC RAID.

I think we had 12 loops, and XLV running across all of them. Each unit was a RAID, and we used XLV to stripe across 12 of them. So, in today's terms, it was a RAID-0 of RAID-5's.

It was part of a filesystem test-bed system. Kind of a how-far-can-we-push-it-before-it-breaks sort of thing.

Sun Baked

Jun 1, 2003, 11:44 PM

Originally posted by illumin8
Do you know how many PCI busses they have? If both 64-bit slots are tied to the same PCI bus then they are sharing that 512Mbytes a second bandwidth.

My point is that Apple seems to be tying more and more system critical functions into the Southbridge. Audio, Gigabit Ethernet, USB, Firewire, and now FC-AL. This would make sense because then the most common peripherals have access to a lot more bandwidth, and your Avid or Protools hardware can have the PCI bandwidth all to itself. The Gigabit Ethernet card is on one PCI channel with the 33 MHz slot. (AGP slot on the Powermac)

The other PCI channel with the Fibre card is on another PCI bus shared with -- the two PCI to ATA bridges (4 Apple drive modules), the low bandwidth I/O USB/Serial/CD-ROM, and the 2 PCI slots where the Fibre card is. (PCI slots and Keylargo on the PowerMac)

The rest of the high bandwidth I/O -- FW, ethernet and such is on the UniNorth chip. (Basically the same as the PowerMac)

---

Apple put the high bandwidth I/O on the Northbridge (UniNorth chip) and the low bandwidth on the Southbridge (KeyLargo chip).

But in the latest PowerMac dropped the USB off the KeyLargo chip -- this was the does my mac have USB 2 discussion.

mcl

Jun 2, 2003, 12:42 AM

Originally posted by illumin8
As always, if you follow the manufacturer's recommendations you will probably be fine. Now that Sun has an enterprise storage offering I highly recommend that you check it out (obligatory plug:) . It is based on Hitachi storage and has been tested to work extremely well with our servers.

At the time, the manufacturer's recommendations (EMC) were, "Only use our GBICs at both ends of the loop. The other manufacturer's recommendation (Sun) was, "Only use our GBIC on the server side."

Of course, I could regale you with horror stories of vendor recommendations, from the time we were asked to keep quiet about the loss of a 1TB array (which was extremely large at the time...consumer hard drives were about 10GB tops) because the vendor supplied us with 1TB of faulty SCSI drives -- the spindle lubricant would suddenly polymerize without warning, causing the platters to seize). And so forth.

These days, my need for storage is more on the order of pushing massive amounts of data over pure fiber links (can only be fiber, because the RF noise must be near-zero) to and from custom inhouse filesystems. However, it must also be done cheaply, so FC-AL's out of the picture (so is anything else that isn't commodity, to offset the cost of the custom boards we must design, build, and install).

BTW, did you work for Entergy? Just curious, I know a couple IT guys that work there.

Nope. ComEd (greater Chicago tri-state area. I believe they still own/run the highest number of nuclear power plants in the country (12?) though the NRC usually has about half of them shut down due to various problems at any given time).