Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Let It Roll

Wow, that was fairly unexpected. There were some rumors that the new Xserve would be coming out in a few weeks, but when I saw Apple today, there were the machines! Wow, how many new products is that in just the past month? Apple did a pretty good job in keeping this release quiet, too.

As for the Xserves get 1.33 GHz chips, it is probably due to the amount of heat the chips give off and how well the units can dissipate the heat (otherwise, it might be an Xblob). I haven't seen an Xserve before, but don't they get pretty loud due to the fans and how they have to push the heat out of the unit?

But certainly nice to see a RAID now. Don't need one, but if I had a business, it would certainly be a nice added tool. Or perhaps even better, get several 1U units, and perform a RAID through those, just in case one of the machines goes down, the others are still there.
 
Originally posted by arnold2
There's more...

Download the spec PDF and have a good look at the detail:

ATA/133 Apple Drive Module (7200 rpm)
(instead of ATA/100)

64-bit, 66MHz PCI slots
(instead of 33MHz)

So, like the first Xserve, this is giving us a taste of the PowerMacs for summer release - the 7457 will run from 867 to 1.833GHz.

Perhaps the NEXT Xserve will have the IBM 970 ?

Naturally;)
 
Originally posted by GPTurismo
So you have 14 drives, and you can only read or write to each one. So the problem there lies that if you can't read from a drive to get the final part of a file or even the first part because it is writing you slow down, who cares you can write or read from all the drives simultaneously if they can only do one at a time...

all i can say is CHEAP LOW END SOLUTION

Oh, no. Another SCSI bigot. Sigh.

Look, IDE's not that bad anymore. From a transfer rate standpoint, it's similar to SCSI; the newest drives for each take turns outrunning each other, but no one really cares once you start doing RAID--even 4 drives will push over 200 MB/sec, and that's too big of a firehose for most small servers to keep up with for any extended period of time.

What really matters, generally, is the number of small, random I/O operations that the drive can perform per second, and that's mostly a function of rotation speed and seek time. Traditionally, SCSI has been quite a ways ahead of IDE on this front, but the lead is dropping. Plus, IDE's painfully cheap price per disk means that you can ususally afford 2-3x as many drives for the same money, which will usually negate SCSI's random I/O lead. Plus, newer IDE drives are adding things like tagged queueing (which the Xraid drives probably support) which should boost IDE drives up to about the same speed as SCSI drives with the same rotational speed. Plus, the next generation of IDE drives will be pushing 10k RPM, the same speed as most SCSI drives and only one notch below the fastest 15k RPM drives.

So, in another year or so, IDE will be approaching drive-for-drive parity with SCSI. Right now, it's slower per drive, but it's so much cheaper that you can probably get better overall performance for the same money.

The half-duplex thing is a complete myth. No currently shipping drive can read and write at the same time. Almost all SCSI drives and some newer IDE drives can accept write requests while read requests are outstanding, and vice-versa, but you don't want to use this feature with writes unless you have a battery backup on your RAID controller, because it will cause corruption on power loss.

That leaves reliability--traditionally SCSI drives have been more reliable then IDE drives. In a decent RAID, though, this doesn't matter much. You just design the system with hot spares and plan for drive failures. Have a cold spare or two around to swap in for failed drives and you should be able to keep running even with an improbably high failure rate. It doesn't matter if you're using IDE or SCSI--if your data matters, then you REALLY need to understand what happens when things start failing, because it's going to happen, it's just a matter of time, and the unemployment list is just as long for SCSI users as IDE :(.
 
Originally posted by Scott Laird

Look, IDE's not that bad anymore. From a transfer rate standpoint, it's similar to SCSI; the newest drives for each take turns outrunning each other, but no one really cares once you start doing RAID--even 4 drives will push over 200 MB/sec, and that's too big of a firehose for most small servers to keep up with for any extended period of time.


Right-o. Also keep in mind several things:

1) Xserve RAID can have up to 1GB of cache. That helps tremendously (What do you think has kept EMC in business all these years?)

2) Fibre Channel provided 200mb/sec *per port* to the Xserve. Being that it has 2 ports, I wonder what it would take to load-balance on the host side (similar to Veritas' DMP or Sun's MPxIO or HPaq's SecurePath)...

What really matters, generally, is the number of small, random I/O operations that the drive can perform per second, and that's mostly a function of rotation speed and seek time. Traditionally, SCSI has been quite a ways ahead of IDE on this front, but the lead is dropping. Plus, IDE's painfully cheap price per disk means that you can ususally afford 2-3x as many drives for the same money, which will usually negate SCSI's random I/O lead.

That doesn't matter nearly as much anymore. First and foremost, if you use this thing with a RAID5 setup, you're nuts. RAID0+1 or even RAID10 is a better option -- RAID5 is just going to hamper performance....

The only place I can see RAID5 being useful is when you only have So Many Disks (tm) to work with --- in something like this, with 2.5TB in 3U of space, there's no need for it ...

Secondly, that big-old-cache in front of the drives is a major help....

What we all need to remember is that the enterprise storage game looks _absolutely nothing_ like desktop storage! Nothing at all!

This is a wholly different beast .... I can't wait to get ours so we can play with it at work ....

I was thinking of hooking up an HP Enterprise Virtual Array (EVA) to an Xserve just to see if it works ... now that apple sells a "supported" Fibre channel card, I can go for it!

--barfoo
 
Originally posted by MrMacman
Does no one care about 10.2.4 ???

AHHH!!

not really, not in a thread that doesn't have anything to do with
Mac OS 10.2.4. Feel free to start another thread if you feel it is
necessary.

These are really nice, I would have liked the rack to be a bit faster,
but I guess it isn't all that bad. A mediocre update if you ask me.
 
IDE drives

From my own personal experience, I've had pretty good luck with IDE drives, they have tended to be the last thing to stick around on some systems I've had. On a junker PC I had, close to everything died on it within two or three years. I think that drive might still work, but it is a 'paltry' 2 GB, so it doesn't do me tons of good in a system now, especially compared up to 60 GB drives.

But on Macs in general, I've had tons better luck with hardware, not just being compatible, but just sticking around and working for awhile. On all of my Macs, the only problems I've ever had have been with CD drives, and that's it.
 
Another scsi bigot?

Look, ATA is fine for workstations and laptops. But for a server environment? Maybe for a small file server. The thing is is that apple is trying to put this up to a high end solution or even a mid range solution and it doesn't cut the mustard.

The main thing is cost, when a G4's biggest advantage is altivec, and that is primarly used in the GUI, why do you need that in a server?

Maybe when they go to a newer 64 bit archetecture, but even then I know of a lot of oracle dba's and other mid range up solution providers that would scoff at the xserve, due to the fact you can get an IBM x330 with dual p3's for $1500 and it will stomp the hell out of that xserve running oracle and/or a web server with linux. :(

Then the EXP 300 is cheaper than the Xraid and it's scsi, sure the drives aren't as large, and are a little more expensive, but for reliability and everything else it's worth it and you will have the extra cash from from the server to buythe extra HDDs.

If you are setting up an internal file server, with under 150 users, and you wanted to maybe host your intranet on it and cash wasn't a problem go Xserve.

But with OSX supporting NFS, and Linux having netatalk and samba, you can get a good linux box cheaper and have all your macs connect to it.

:(

Reality sucks :(
 
ATA???

Forget the current technology of ATA/133 or 150, I'm really hoping for SERIAL ATA which begins at 150MB/s up to 600MB/second data transfer for read/write internally in the drive itself.

Now If i'm not mistaking isnt the SCSI technology capable of reaching 360MB/second? Either in the LVD or Ultra LVD or another spec??? And no I'm not just talking about whats currently available for the MAC but also for PC??

Glad about the XRaid and the fibre channel.....is the fibre channel really just the FireWire2 port connections???
Can somebody post a pic on the actually connectors?

Peace.
 
Re: ATA???

Originally posted by Prom1
Forget the current technology of ATA/133 or 150, I'm really hoping for SERIAL ATA which begins at 150MB/s up to 600MB/second data transfer for read/write internally in the drive itself.

Doesn't matter. Even Ultra 66 is fast enough for a single drive, and Ultra 100 or 133 is fast enough for another year or two. There are three reasons for this:

1. The fastest drives are around 60 MB/sec right now.

2. 4 drives at 60 MB/sec are faster then 200 MB/sec FC.

3. The only common application that needs I/O rates this high is uncompressed HDTV editing/capture. Remember, it'll take multiple gigabit ethernet pipes to feed 200 MB/sec to a server. In real life, random I/O almost always matters more.

The real-life performance difference between 14 drives, each via a 60 MB/sec connection and 14 drives, each via a 600 MB/sec connection is freakishly close to 0.

Having said that, Serial ATA has a number of nice things going for it (smaller cables and connectors, standard support for hot-swap), but 150 vs 300 vs 600 is irrelevant for single drives in a RAID array.


Now If i'm not mistaking isnt the SCSI technology capable of reaching 360MB/second? Either in the LVD or Ultra LVD or another spec??? And no I'm not just talking about whats currently available for the MAC but also for PC??

Yep, Ultra320 is currently the fastest standard. Remember two things, though:

1. You almost always put more then one drive on each SCSI bus, so they have to share.

2. SCSI is frequently used to connect hosts and RAID arrays, so higher performance actually matters at lot more then performance on single IDE channels.

Having said that, a single Ultra320 channel pretty much needs a 64/66 PCI bus to itself if you plan on getting full speed out of it.


Glad about the XRaid and the fibre channel.....is the fibre channel really just the FireWire2 port connections???

Totally unrelated. The standard copper FC connector looks kind of like a mutant RJ-45 as I recall. Older copper connectors typically used DB-9 connectors, although it's actually possible to run FC over coax, if you have the right coax.
 
Relibility

"because it's going to happen, it's just a matter of time, and the unemployment list is just as long for SCSI users as IDE ."

Have you all forgot about S.M.A.R.T

All the ide drives support this and it's what makes them more relible now then a few years back, it monitors the condition of the drive and gives you realtime feed back about the disk, so you pull it out before it dies giving you 100% protection becouse it's be replaced before it dies in your server!

at a place i know that uses servers the have about 5 hp servers with about 600gigs a server they still have a tape drive to back them up and they are like 60 pounds not including the UPS power that only last for like 20 min, and they way as much or more!, they have there own airconed room that mesures like 14x14ft! with apples new xserve you could fit it in a closet!

As for speed what more do you want it to do cook your dinner!

:)
 
Details

Details ppl.

Fibre Channel for use with the PowerMac G4? No I say, Why??

If you read the PDF's for the XServe Raid and the XServe, and the PowerMac G4 (1.42Ghz or any others). You'll see what I'm talking about.

Fibre Channel PCI card requires 64/32-bit 66Mhz PCI slot according to the XServeRaid PDF (Feb 2003); which both the XServes have, but the PowerMac G4's, ALL in existence, only have 32-bit/64-bit (64-bit for the 1.42Ghz model) 33Mhz PCI slots. On page 3 of the XServe RAID pdf under the heading Apple Fibre Channel PCI Card (sold seperately), it states 64-bit 66Mhz dual SFP 2Gb Fibre Channel ports compatible with 32-bit 66Mhz PCI slots. (order #8940G/A).

So what gives with the picture of the PowerMac using a RAID???

ScottLaird,
Thanks for the info but in setting up RAID drives 1 and up I do believe random access for hosting or even applications for Excel databases--huge sheets of info......ATA 100/133 would benefit even the average user. point in case is the XServe actually using them. Serial ATA which is being built by Seagate, not sure whom makes the PCI card for them actually would help the consumer but not by leaps and bounds I agree.

Cheers
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.