Xserve Competes With High-End Unix Servers
wayneh writes "There is a great article at ITworld.com about how Apple's Xserve is finding its way to high-end server vendors. The vendors who traditionally sold Sun and IBM servers are now looking into and stocking the Xserve as their clients become curious about the system. It'll be interesting to see how well the Xserve does among its more traditional competitors."
It'll be nice to see Apple get out of its rut as a graphics machine and only in schools. Macs are great machines... it used to just be the OS that (tech) people didn't like... and now even thats not a problem. (Did I get the first post? :)
I'm a diehard Mac guy, but I'll be the first to admit that Apple has not put out competitive servers before the XServe. When Apple changed from offering basically souped up Powermacs with a non *nix OS to one of the best 1U servers on the market running OS X, does it really surprise anyone that they are going to be getting attention in markets of which they traditionally were not even on the radar?
This is great to see. I hope that Apple can scale their production volume to keep up with the demand. I think one of the major selling points is that it comes with an Unlimited client license for Mac OS X Server, unlike any Windows 2000 Advanced Server setup. Licenses are expensive, and I know that's been a major factor in us moving away from Novell NetWare here at my university.
If you remember way back in May when Apple introduced the Xserve, they also previewed a 2GB RAID solution. According to this relatively old c|net article:
Apple also previewed a future storage device, the Xserve RAID, a 5.25-inch thick cabinet that can contain 14 hard drives for a total capacity of 1.68 terabytes. The system has two 2-gigabit-per-second Fibre Channel connections, a high-speed connection technology for communicating with servers.
There have been some rumblings around the Mac rumor community that this will soon debut. Can I get a "booyah"?
-- My hovercraft is full of eels.
I'm actually waiting for a faster processor (scaled down Power4, etc.) and better memory bandwidth. I would like a 64 processor rack for some serious computations. Though I can afford it, I don't want to go down headache-road with a PC-based rack. Does anyone have an Xserve with myrinet?
At the end of the article, it says:
"The challenge is, who is going to buy it?" Eunice said. "There is so much pricing pressure and competition in the market. The reality is that Apple will have a hard time going to financial communities or telcos with this product."
Apple gives you an UNLIMITED client license - how can the article offer this as a serious concern when licensing cost is such a big concern, especially for Micrsoft houses?
As I recall, IBM doesn't have a license to run the AltiVec stuff, which is kind of a hinderance to deploy machines the IBM Power processors, but maybe that doesn't affect the Xserve machines. I wonder if they'll finally get a license or resolve that. I could be wrong though.
Erm... I'm confused.
I thought that the Apple Xserve boxes were 1U Dual 1GHz G4 machines with ATA/100 7200rpm Hard drives.
How exactly does this compare with the type of High-End systems that companies like IBM and Sun sell?
These machines rank along side the Entry Level / Workgroup type machines and are hardly the cutting edge of power.
I like Apples kit, but please call it what it is.
I've only read about the Xserve, and have not heard it pronounced by Apple.
Is it pronounced:
X-serve?
Ten-serve?
Zerve?
__ Someday, but not this morning, I'll finally learn to use the preview button.
This is a completely baseless assertion. The constant may not be 4. Clustering may not be suitable in a specific situation. And you haven't remotely touched on reliability, support, or any other decision criteria-- many of which are far more important to TCO than initial purchase price of hardware and OS.
I do not have a signature
The current issue of the german c't magazine has a comparison of Apples XServe vs. Dells PowerEdge 1650.
Basically what they are saying is that for low-end office serving the XServe does well, however, as more performance is needed, the PowerEdge clearly wins (Linux). Administration and integration in Mac OS X environments are strengths of the XServe. Want more power, go for Dell/Linux.
Lars
In the article they talk repeatedly about the ability of XServe to talk with Sun boxes. They also talk about the XServe filling a niche Sun doesn't.
Is it me or would an Apple/Sun alliance make a lot of sense? I mean, besides the egos involved. You'd have server (high/low end and database) coupled with desktop.
Plus you'd have all of the stuff that MicroSoft wants working together (clean desktop for idiots, server market, stability, security) Just wondering
The opposite of progress is congress
Unix + x86 + clustering = 40 times the admin/headaches of Apple.
That was classic intercourse!
I still don't know how well apple will do with xserve - but I do know it has some great possibilities. My thoughts are that apple snuck this one past them all ... and will hit them hard. Unix + sleek hardware + sleek front-end ... go apple!
mx
There must be some truth in Apple's benchmark http://www.apple.com/xserve/performance.html BLAST results: At the short word length of 9, the Xserve is 21 times faster than the IBM eServer x330 and 52 times faster than the Sun Fire V100. At the long word length of 40, Xserve is 5.8 times faster than the IBM eServer x330 and 13.4 times faster than the Sun Fire V100.
The XServe is already fairly cheap for the power it provides. The ability to put 480 gigs in the box only adds to this cost-effectiveness, given the cost of rack space (though they need to make the internal drives SCSI). The company I work at has been very successful using lots of inexpensive boxes to scale to a large number of users. I think something like the XServe would be excellent in this type of environment.
However, Apple needs to decide where they want to go with the XServe and their server business in general. Do they keep it as an entry-level offering and add bigger servers above it? Do they build a clustering system to make it easy to scale in the way I mentioned earlier? Both? My fear is that they will try to target it to their traditional core customers (artists, designers, etc.) rather than using it as an opportunity to branch out into new areas. In other words, they'll make it work really well as a rendering box but not as a server. On the other hand, one could certainly question whether the world needs another UN*X server vendor and whether Apple is equipped to take on Sun, Dell, and IBM.
On a different note, what about an Apple X (well, Aqua really) Terminal? A lab full of Apple Terminals powered by a half-dozen XServes racked in the corner would be pretty cool...
If the plural of octopus is octopi and the plural of server is servi. Is the plural of bus bi? The plural of us must be I.
I have a second sig, I call it sig#2.
It sounds to me that you are biased and not capable of making rational decisions based on facts.
First of all, Darwin is completely open source, you can compile the kernel on both Mac and PC, and Apple don't force anyone to use their GUI . Secondly, what has MS got to do with Apple? The last (and the only) OS X security issue was fixed by Apple within 3 days since the discovery.
The problem with an open source hothead like you is that your argument is driven by ideology and emotion rather than logic, and it's really annoying.
>> Apple can put backdoors in their GUI or iTunes or whatever as easily as the kernel. If the whole system isn't all open, you'll never know FOR SURE.
Is there evidence or reason for Apple to do this, or is this just total garbage from your otherwise empty head?
>> I'll stick to Free/OpenBSD/Gentoo/Whatever before I go to a kernel that's obsolete and badly designed but hanging on by pure ca$h infusions from Apple and the macaddicts who would buy a bag of crap if Apple put their logo on it.
Obsolete and badly designed, by what standard? Have you ever used OS X before you open your filthy mouth? I would say Apple's OS technology is at least 5 years ahead of Linux or anything else in the industry.
>> Hell, why would they choose a Mac over Win2000? I wouldn't, and I hate Microsoft.
Xserve with unlimited client license is priced similarly to a Dell with a MS server OS for 10 clients, not to mention OS X's far superior stability and security.
>> If Apple was bigger, it would be AS BAD or WORSE than M$.
How do you work that one out?
>> If you'd ever adminned an NT system or owned a Pinto, you'd understand. Apple's offering is not a very strong contender to get ME to buy.
You sound like an MSCE pretending to know something about open source.
What is the cost of multiple Linux licences?
FreeBSD?
NetBSD?
What is the cost of OS X Server?
Now double the Number of machines that need the licences to achieve the same performance that x86 offers...
If voting were effective, it would be illegal by now.
The cost of OS X Server when you buy an XServe is exactly the same as the cost of multiple Linux/FreeBSD/NetBSD licenses: $0. How do you like them apples? (pun intended)
The Xserve is a LONG beast, easily the length of a typical rack, where most 2 or 3U servers are half that length.
Generally, the Xserve is a sweet beast in speed and performance, particularly with a prerelease of 10.2 Server I have installed.
Some of Apple's claims are weak right now, although they are doing a bit to help me with that now. The biggest disappointment was LDAP/Windows Active Directory authentication, which failed miserably in my 10.1 tests. My 10.2 update may have cleared that up, but Apple's documentation group needs some infusion from the other server OS documentation people for more concise instructions.
It's support apps are very good, and the OS sticks to virtually all IP standards, making the thing easy to administer. Configure, no, but administer, yes.
I can see this box being a good, less expensive alternative to a few of the Compaq boxes sitting around it.
The real gem of the Xserve is not the box, but the power of the OS behind it. This box would not be possible without Mac OS X Server 10.2. I have 10.1 Server running on an older 2-processor G4, serving a heavy load. It is a very stable, efficient box.
Vos teneo officium eram periculosus ut vos recipero is.
If you can build so many IU Servers for this price, provide the links to all the component parts and a proper calculation of its costs or stay out of the debate.
For $2,999.00 DIY you can buy THREE systems consisting of Dual Athlon 2000 @ 1.67Ghz, 120 gig drive,1 Gig ram and dual gigabit ethernet (2 will fit in 1u)
Where can I order these? I'm interested.
For whatever it's worth someone's reporting that the IBM vectoring is NOT Altivec and isn't Altivec compatible.
I vote for calling it Altivec2 and using this chip whatever it takes... I'd much rather hitch my future to IBMs engineering and manufacturing than Moto's.
=tkk
Bill Gates - Creationist?!?
I'm sure someone else has already said this but here it goes anyway:
Apple, Motorola and IBM all created the 160 128Bit instructions cooperatively. Altivec is the brand name that Motorola uses, it has nothing to do with the instruction set. IBM can use the instructions they just cannot name it Altivec unless Motorola says okay and allows them to license the name or just use it for free (unlikely).
So IBM is integrating the "Altivec" instructions it will be branded a different name more than likely.
Seriously, I use macs for all my desktops, but since I've seen the JFS light on linux, I won't even consider running a server without a journaling filesystem. HFS+ is far too fragile - I had to reformat both my G4 and my powerbook to upgrade to Jaguar (fsck.hfs+ doesn't fix all hfs+ problems; it reports them and tells you to go find a better disk utility) and UFS doesn't have a journal.
This is marginal at best for a desktop and totally unacceptable for a server. Apple can't play in this market if they're not willing to cover some really basic software requirements. They've got some great hardware in X-Serve, but who's going to want their big RAID array if you can't store files reliably on it? They need to move beyond thinking, 'oh, we'll get lots of QTSS customers using them' if they really want to make inroads into the market.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
RAID and Journaling Filesystems aren't interchangable, they solve two separate problems. They're complementary technologies.
RAID provides for hardware reliability. Lose 1 disk, and you don't lose your data.
A Journaling Filesystem provides for filesystem integrity in the event of a catastrophic event (complete loss of power, kernel panic, system hardware failure, etc.). Roughly, data is written to a journal, the journal is marked dirty, the data is copied to the filesystem, the journal is maked clean (I know folks, I'm oversimplifying).
With a straight RAID solution if an event occurs that would corrupt your filesystem, say a system crash, you have two or more copies of the same error, which doesn't help. With a journaling filesystem, the journal is replayed at fsck time, and any interrupted data writes are stored onto the filesystem safely.
This method is extremely important for real servers because fsck can be skipped on reboot if need be (even if it's not ideal). fsck can take an unacceptably long time on sufficiently huge volumes (I run a couple 2 TB volumes, it's deadly). You could never *think* of skipping fsck on a non-journaled filesystem. The best implementations will store the journal in NVRAM and/or have disks without write buffering.
Apple's fibre channel RAID system looks like great hardware, but without a journaling filesystem it's going to take time on the order of hours to come back up after a crash with a full X-serve cabinet. Noone can run a mission-critical server like that.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)