1U And 3U G5 Xserve's?

[mstecker]

3U

The 3U may be a cheaper altertnative, as I suspect that it could use the stock G5 motherboard.

Just a thought.

Matthew

 

[hmmfe]

Re: 3U

Originally posted by mstecker
The 3U may be a cheaper altertnative, as I suspect that it could use the stock G5 motherboard.

Just a thought.

Matthew

I'm not sure I'd go with the stock G5 mobo, but I see your point. Of equal importance, end user upgrades are generally easier and cheaper too.

 

[ffakr]

3U not a good idea for clustering

Apple has a focus on Clustering right now. They are doing something with xGrid.. they got a lot of acclaim for the Big Mac, they did a session on clustering at WWDC last year...

3U servers are not what the market demands for clustering.

3U servers are always designed for larger internal drive arrays. Apple would only release a 3U server if they felt that the market needed a Mac server with a lot of internal drive space, or a large SCSI array.

Apple won't release a 3U for the following reasons...

- Apple has proven that they can deliver high speed and high capacity from a 1U server running an IDE drive subsystem. With 300 GB IDE drives, Apple could easily deliver around 1TB internal with SATA.. and well over 1 TB in a 2U.
- Apple has a product for large storage capacity. xServe RAID. They have laid out their enterprise strategy for now and it's xServe w/ xServe RAID for large capacity storage
- 3U servers cost $$. They typically used for 4 way boxes, or boxes with larger internal disk arrays (quite possibly SCSI for DB servers). Apple doesn't yet have the enterprise credentials to tackle these markets.. People whose jobs rely on uptime won't put it all on the line to dump $10,000, $20,000 or more on a Mac... Not yet. ;-)

Apple is sneaking into the enterprise. They've got two ways in... sneak in with lower priced xServes and sneak into low budget Edu environments. The goal is to change this and make Apple an enterprise player, but that will take time.
I don't see an expensive 3U server fitting into Apple's markets right now.

I'd bet my mac on it (it's a 6100/66 so I don't really mind being wrong)

 

[hmmfe]

Re: 3U not a good idea for clustering

Originally posted by ffakr
Apple has a focus on Clustering right now. They are doing something with xGrid.. they got a lot of acclaim for the Big Mac, they did a session on clustering at WWDC last year...

3U servers are not what the market demands for clustering.

3U servers are always designed for larger internal drive arrays. Apple would only release a 3U server if they felt that the market needed a Mac server with a lot of internal drive space, or a large SCSI array.

Apple won't release a 3U for the following reasons...

- Apple has proven that they can deliver high speed and high capacity from a 1U server running an IDE drive subsystem. With 300 GB IDE drives, Apple could easily deliver around 1TB internal with SATA.. and well over 1 TB in a 2U.
- Apple has a product for large storage capacity. xServe RAID. They have laid out their enterprise strategy for now and it's xServe w/ xServe RAID for large capacity storage
- 3U servers cost $$. They typically used for 4 way boxes, or boxes with larger internal disk arrays (quite possibly SCSI for DB servers). Apple doesn't yet have the enterprise credentials to tackle these markets.. People whose jobs rely on uptime won't put it all on the line to dump $10,000, $20,000 or more on a Mac... Not yet. ;-)

Apple is sneaking into the enterprise. They've got two ways in... sneak in with lower priced xServes and sneak into low budget Edu environments. The goal is to change this and make Apple an enterprise player, but that will take time.
I don't see an expensive 3U server fitting into Apple's markets right now.

I'd bet my mac on it (it's a 6100/66 so I don't really mind being wrong)

Well thought out opinion and you are probably right on the money as far as what Apple will do.

I disagree with the strategy though. I would think that Apple would target the SMB market with a low-cost user upgradeable 2U-3U server. There is nothing inherent in a 2U or 3U chassis that makes it expensive. In fact, they are generally cheaper when you figure in disk space and the use of OTS upgrades.

I've been selling and installing servers in the SMB market for years and almost never encounter a scenario where a 1U server makes more sense. That is not to say that I never use one - just that the main effeciency (vertical space) is not a concern when you may have 3 or 4 devices in a relay rack stuck in a broom closet. And speaking of which, some of these racks were installed with 12-18 networking gear in mind and are very cramped with the 28" XServer sticking out back. It would be nice if everyone had a nice new 35" deep cabinet to install stuff into but usually they are bare relay racks cramed into a utility closet.

If you are right about Apple going after the cluster market (and presuming they will not try to enter another market with a different server product), then a 1U server or a blade chassis is the only way to go.

Either way, Apple will stick with the 1U design because - if nothing else - it is a nicer design (i.e. sexier). That being said, I hope they come out with a slightly larger more functional chassis.

 

[nek]

I don't have any need for a server, so I don't know a lot about them, but I also don't understand this 3U Xserve speculation. I think it is much more likely that the Xserve will remain 1U and the rumored 3U item is just the Xserve RAID.

IBM's BladeCenter JS20 comes with dual 1.6GHz PPC 970, so why wouldn't Apple create a 1U Xserve G5?

 

[ktlx]

Originally posted by Frobozz
Technically, I see your point and agree... but I got thinking... if Memory bandwidth is such an issue then how does a cluster work sucessfully? Is it because each set of processors has it's own set of memory bandwidth, and the communication between machines is restricted by the throughput of the cabling?

Clusters are successful only on a set of problems that can be easily decompossed into chunks that require a lot of processing compared to the amount of communication bandwidth. Fortunately there are quite a number of important scientific problems that fit this model well.

However they blow for things requiring lots of I/O. That is why you see clusters used for fluid flow and nuclear reaction simulations while large monoliths used for databases and transaction processing.

In my opinion, that is the weakness with Apple's so-called "Enterprise Strategy". Most enterprises need things like file, print, Web, database and transaction processing servers. Very few actually do nuclear reaction simulations or render 3D animations. A dual processor 1U server is fine for file, print and Web services but everyone and their brother already meets the needs of that market. A cluster of dual processor 1U servers is not really the appropriate solution for database and transaction processing.

 

[ffakr]

Re: Re: 3U not a good idea for clustering

Originally posted by hmmfe
Well thought out opinion and you are probably right on the money as far as what Apple will do.

I disagree with the strategy though. I would think that Apple would target the SMB market with a low-cost user upgradeable 2U-3U server. There is nothing inherent in a 2U or 3U chassis that makes it expensive. In fact, they are generally cheaper when you figure in disk space and the use of OTS upgrades.

I think that windows filesharing is an extremely risky market to dabble in though.
First off, you have to bump up against the MS bias against Macs.
Second, though Apple can integrate into AD and it can even act as a PDC but you won't see an AD running ON a Mac OS X Server. the truth is, Windows Server just integrates better into an AD and AD is where it's at for anything but small shops.. if not now, than in the near future.
I think you're getting at small installations where linux is an option but Linux still has more street creds than Mac OS Server (though I prefer FreeBSD over linux for a server). In this case, small businesses that would consider Linux smb servers, Apple has a shot.. if not now then soon. I just don't think it warrants the hardware investment yet.

I'd love to see Apple do well enough that they had 1U servers, and 2-3U file servers, and 3+U quad-proc database servers but I just don't think they are there yet. Apple can't splinter the small market is always has, it should focus on where it's doing well and try to work on enterprise creditability.

But as always, I have to point out that Apple doesn't seem to value my strategic opinion to seek me out and pay me fat $$. ;-)

BTW.. I even think that Apple should buy SGI and make low volume/high margin big bad supercomputers running Mac OS Server. How about an Altix with 980s or Power5s... mmnnn good. If you think Big Mac is good for business, imagine if Apple became the premier hardware and development enviornment vendor for High Performance Computing. Hey, it could happen.
[/B][/QUOTE]

 

[ffakr]

Originally posted by ktlx
In my opinion, that is the weakness with Apple's so-called "Enterprise Strategy". Most enterprises need things like file, print, Web, database and transaction processing servers. Very few actually do nuclear reaction simulations or render 3D animations. A dual processor 1U server is fine for file, print and Web services but everyone and their brother already meets the needs of that market. A cluster of dual processor 1U servers is not really the appropriate solution for database and transaction processing.

Well, a modern dual processor server is plenty powerful enough for a lot of database needs. The databases that require quad, 8, 16, 32 proc machines also come with big IT budgets for hardware. It is totally one thing to get a Mac server into a server room full of Dells and Suns if you are talking a $3-5000 machine. It's really different when you are talking about passing up the $50,000 Dell or Sun Database server in favor of a big 8 way Macintosh. I just don't see that happening.

I agree, Apple is missing out on a whole segment of the market, but is it a segment that they can sell machines in at this point?

IMHO, the biggest limitation of Apple's current server option is the storage. Apple's proven that IDE based storage can be pretty fast compared to other 1U IDE and SCSI servers. Unfortuantely the current configuration is pretty crappy. The current xServe only supports 4 single volumes, or RAID 0 and/or RAID 1. You can't have RAID 5 volumes that span all 4 drives. You can't have one large volume.

I think Apple should stick to 1U. They should provide dual G5s over 2GHz with room for at least 8 DIMMs (up to 16GB of ram support with 2GB DIMMs). They should provide real RAID 5 hardware support with SATA drives up to 250GBs. They should offer 10K SATA IDE drives OR offer a Ult320 SCSI drive option instead of IDE.
It would be a fast fast machine that could chew through fairly large databases yet it would remain small with a low $$ entry point. It would, however, provide the first step to a mid range server.

Apple should follow with a 3 or 4U down the road. It should be configurable with 2-4 G5s (or 980s). It should take up to 32GB of RAM. It should take 5-6 IDE or SATA drives in RAID 5.

I also think that xServe RAID should be upgradeable to be a stand alone SAN. But that's another can of worms.

 

[hmmfe]

Re: Re: Re: 3U not a good idea for clustering

Originally posted by ffakr
I think that windows filesharing is an extremely risky market to dabble in though.
First off, you have to bump up against the MS bias against Macs.
Second, though Apple can integrate into AD and it can even act as a PDC but you won't see an AD running ON a Mac OS X Server. the truth is, Windows Server just integrates better into an AD and AD is where it's at for anything but small shops.. if not now, than in the near future.
I think you're getting at small installations where linux is an option but Linux still has more street creds than Mac OS Server (though I prefer FreeBSD over linux for a server). In this case, small businesses that would consider Linux smb servers, Apple has a shot.. if not now then soon. I just don't think it warrants the hardware investment yet.

I'd love to see Apple do well enough that they had 1U servers, and 2-3U file servers, and 3+U quad-proc database servers but I just don't think they are there yet. Apple can't splinter the small market is always has, it should focus on where it's doing well and try to work on enterprise creditability.

But as always, I have to point out that Apple doesn't seem to value my strategic opinion to seek me out and pay me fat $$. ;-)

BTW.. I even think that Apple should buy SGI and make low volume/high margin big bad supercomputers running Mac OS Server. How about an Altix with 980s or Power5s... mmnnn good. If you think Big Mac is good for business, imagine if Apple became the premier hardware and development enviornment vendor for High Performance Computing. Hey, it could happen.

[/B][/QUOTE]

Well, I was not envisioning a Mac Server driving Windows desktops. Rather, I was envisioning Apple coming out with a solution that stands on its own. Admittedly, it is a small market - or even non-existent maket at the moment - but I think if Apple is gonna be taken seriously at all they need to deliver an all-round solution.

Linux. I sell and install a bit of that as well. What I'm suggesting is that if Linux can compete in this space so can (and should) Apple.

Anyway, all fantasy stuff... what I would like to see. Good chatting with you!

 

[ffakr]

Originally posted by ffakr

I think Apple should stick to 1U. They should provide dual G5s over 2GHz with room for at least 8 DIMMs (up to 16GB of ram support with 2GB DIMMs). They should provide real RAID 5 hardware support with SATA drives up to 250GBs. They should offer 10K SATA IDE drives OR offer a Ult320 SCSI drive option instead of IDE.

OK, not bad.

Apple ships dual 2GHz with SATA drives in sizes up to 250GB and they provide hardware RAID 5 support with an optional PCI card.
They support up to 8GB of RAM with 8 DIMM slots but I assume that, like the U3 on the desktop, it should take up to 16GB when and if Apple certifies the use of 2GB ram sticks.

 

[AlanAudio]

With the speculation over 3U servers, we can see a number of theories being suggested here.

The first is that you need 3u for a huge array of hard drives.

Personally, I don't buy that, as hard drive technology is moving quickly, so capacity and speed are improving very rapidly and those who need to go beyond what will fit into a 1u case would be better using a 3u rack of drives anyway.

The second is that it would make an awesome powerhouse for video editing, ProTools( Audio recoding ) and their multiple add-on cards.

Much as I would like that to be true, I don't see it, as Apple are more likely to market a Pro-Tools rival, just as they did with Final Cut Pro in the video world, than to make life easier for a company ( Avid ) who favour PCs.

The third theory is that 3u boxes would be cheap to make as you could re-use existing motherboards.

Since when was Apple a producer of cheap stuff ? They make stuff that is cost-effective, but doing things just because they are cheap doesn't sound like the Apple I know.

The fourth theory is that Apple may be planning a blade-type server - I would put my money on that theory.

The concept of something like 8 G5s in a 3u case does seem to fit in with Apple's way of doing things.

When you add that to their recent amazingly low-key announcement of X-Grid, which allows multiple Macs to work co-operatively to become one supercomputer, it only adds weight to that proposition.

It may be that the existing mass business market for servers will not be properly addressed by this product, but I would imagine that Apple has identified an entirely new market which would need such a product.

With Steve's involvement in Pixar, the 3D animation business is an obvious contender, together with supercomputers for scientific research, but I bet he has something else in mind as well.

 

[ogminlo]

With that announcement that Apple has a 26% market share of the video production market, I would not be surprised to see the next round of G5s include a rackmount chassis option. The Current G5 chassis will not lay sideways in a standard 19" rack, and a 3U Rack G5 would allow for three to live where now just two can with wasted space around them. There is room for the current innards to fit if you remove the handles. No other changes would be necessary! I think this is a real option for Apple and this 3U Xserve rumor just missed the market Apple was building for.

I know our shop would love to rack our new G5s properly. With such a strong market share that would appreciate a rackmount option, I'd say this is viable.