Enterprise Mac | Tom Yager » December 2006

December 31, 2006 | Comments: (0)

Quick Review: 15-inch Core 2 Duo MacBook Pro

TAGS: Desktops and notebooks

InfoWorld awarded the Core 2 Duo MacBook Pro its Notebook of the Year for 2006. While I wrote of it in my blog and column, I never penned an official review specifically for Apple's new notebook, relying on my April 2006 review of the Core Duo MacBook Pro with the standard PC model bump presumption ("the previous model, only faster"), but the Core 2 Duo cut of MacBook Pro deserves its own review. MacBook Pro has evolved in a few ways that deserve note. Apple also failed to address the flaws that I've noted in the keyboards and touchpads in this and recent Mac notebooks, but in the end, Apple pulled off a market-beater of a 64-bit commercial notebook.

Whether you're a current Mac user or not, the 15-inch Core 2 Duo MacBook Pro is the notebook you want to carry. Its desktop replacement feature set is ideally matched to a comfortable carrying weight that otherwise qualifies Core 2 Duo MacBook Pro as a mid-class commercial notebook. It outclasses mid-class x86 notebooks in all other regards, with the standouts being 400 and 800 Mbps FireWire ports, a full-sized DVI video port for uncompromising digital output from MacBook Pro's 256 MB AMD/ATI X1600 graphics processing unit, a brilliant display with a nice stiff hinge, a backlit keyboard and expandability to 3 GB of RAM. And then there's OS X Tiger, a stellar application platform in its own right, Microsoft's inspiration for Vista and yet Vista's better in every measure but market share.

The greatest shortcoming that Core Duo MacBook Pro suffered, short battery life, is entirely resolved in Core 2 Duo MacBook Pro. Apple managed to add 30 to 60 minutes to MacBook Pro's running time on battery, with the size of the extra time slice determined by the portion of time you spend with the AirPort wireless transceiver turned on. I must point out as well that the Core Duo notebook I tested before this new model was a 17-inch machine with a 7,200 RPM hard drive, making it a greater consumer of current on at least a couple of scales. However, I exercised due diligence in testing Core Duo MacBook Pro's battery life by running it at minimum brightness and with the hard disk timeout set to 1 minute. That unusual degree of frugality netted me less than three hours of running time, while I routinely pull four hours of continuous computing (no dimming, no disk spin-down, but no wireless) on the Core 2 Duo MacBook Pro.

Subjectively, heat is somewhat less of an issue as well. That's very hard to quantify; the 17-inch Core Duo notebook's more expansive speaker grilles and palm rests created a large hot zone that ran from the Mag Safe magnetically latched charter port and radiated down to the left palm rest. Core 2 Duo MacBook Pro can still get very warm, especially when operated on its charger (which Apple and I recommend to extend the total life expectancy of the battery). But the hot zone seems to be pushed north, concentrated at the grille concealed below the display. I have two other notebooks in Core 2 Duo's performance class here, and both run much cooler, but at the expense of running their cooling fans very liberally. One of the competing PC notebooks also has fan intake grilles on the bottom surface, a common design trait and an insidious enemy of quiet operation.

The performance of the dual-core, 64-bit Core 2 Duo CPU built into MacBook Pro needs no description, except to say that the dramatic perceived performance boost over PowerBook that some reviewers had attributed to Core Duo MacBook Pro is actually delivered by Core 2 Duo MacBook Pro. Applications like Microsoft Office that run under Apple's Rosetta PowerPC translator have, in my opinion, finally achieved an acceptable level of performance without dragging the performance of native applications. This is credited not only to Apple's new notebook CPU, but to Apple's frequent updates to Rosetta itself.

The one significant flaw of Core 2 Duo MacBook Pro is one that actually traces back to late-model PowerBooks. The last two models of PowerBook that I tested, along with two models of MacBook Pro, have had quality issues with both their keyboards and touchpads. With keyboards, I've had keycaps pop off, keys that go dead when the machine heats up, repeating keys, a squeaky and unbalanced spacebar and, with this latest model, a Delete key that makes one time in three or four presses. Touchpads on all of these models have intermittently wandered and clicked themselves when I rested my palms below the keyboard, a problem so pronounced in Core 2 Duo MacBook Pro that I've had to turn off tap-to-click. These strike me as problems that would be caught in the most cursory QA tests. They might not stand out so glaringly if Apple hadn't set the bar for excellence in human interface devices.

I offered Apple a chance to swap out the MacBook Pro review unit for the keyboard and touchpad problems. Apple did not respond, but I submitted my request a bit too close to Christmas.

An honest review of Core 2 Duo MacBook Pro requires criticism for those ways in which Apple fell short of the exceptional notebook standards it set for the entire industry. But most prospective buyers of MacBook Pro have never laid hands on a PowerBook. Even in a hurried retail setting, any side-by-side comparison pitting Apple's Core 2 Duo MacBook Pro against competing Windows notebooks with similar specifications consistently leaves one with the conviction that while Apple's competitors can churn out interchangeable commodity (I meant to type "standardized" but my fingers wouldn't let me) notebooks from off-the-rack innards, Apple's from-scratch design resulted in a notebook that outclasses the mainstream. And if you need any more convincing than that, consider that Core 2 Duo MacBook Pro is Leopard-ready. Being prepped to run Apple's next-generation OS and application platform is far more meaningful and worthy of investment than any PC bearing the "Vista-ready" tag (which was all of them). Everything that Apple has made since the G4 PowerBooks will run Leopard, but Core 2 Duo MacBook Pro will make the best possible mobile living quarters for the remade Mac software platform.

Postscript: Apple is now selling an in-flight DC adapter to address the fact that at 85 watts, the standard AC-to-DC charger exceeds the 75 watt limit of under-seat power outlets. Also, touchpad difficulties are eliminated with the use of Apple's optional wireless Mighty Mouse, which automatically links to MacBook Pro's built-in BlueTooth transceiver. I recommend both to road warriors.

Posted by Tom Yager on December 31, 2006 06:33 PM

December 21, 2006 | Comments: (0)

Kill two Windows servers with one Xserve

TAGS: Servers and storage, Software

At its core, Xserve is a two-socket Core microarchitecture Xeon (Woodcrest) rack server. As I wrote in my review, in hardware design, Xserve lives up to market standards. Some touches, like the SAS/SATA drive bays, a PCI-X slot for existing expansion cards and the SuperDrive dual-layer DVD burner, help tip the scale in Xserve's favor. But the reason to buy Xserve is OS X Server: No other server app platform rivals it, and no other server system runs it. If you want OS X Server, you need a Mac, and Xserve is the only Mac that's equipped with external drive bays and a baseboard management controller.

It is an absolute sin to let any of that firepower and capacity go to waste. There are all kinds of ways to shuffle and reconfigure for load balancing and fail-over. If you play it right, bringing in a new server means knocking out an older, slower server. Maybe more than one. Or buying a faster server in the first place obviates the need for the second or third server that would have been required based on old school rules of thumb.

A constant trend of doing ever more work while reducing the number of power supplies used to do it is my definition of consolidation.

Virtualization is a key enabler in consolidation because migration is done on a whole-server level, not at the granularity of application, configuration or content. In the ideal, a physical-to-virtual (P2V) migration is a blip on users', administrators' and partners' radars. The old server machine dies, and its virtual clone rises up, indistinguishable from the box that you just offed. Not one hair is out of place and neither users nor other servers need to change their ways. That ideal is attainable for most existing Windows Server users; Microsoft gives its virtualization software away. It altered its license policy to allow one authorized key to cover multiple virtual machines on the same host. Say what you will about Microsoft and Windows, but Redmond has done the right thing on the virtualization tip so far.

However, Xserve owners miss out. The Mac platform won't virtualize in the traditional sense of hosting a virtual instance of itself. That's a real pity, especially for those of us holding last rites for PowerPC servers. But there is a silver lining: Xserve is now an x86 rack server. It can't take on the distinct personalities of two OS X Server machines, but it can show Windows and Linux servers the door. My focus is Windows. The specific enabler for using Xserve to consolidate Windows Servers comes from Parallels. Parallels Desktop is best known as the $80 software add-on that allows you to run Windows XP or Vista on your Intel Mac. Fewer know that Parallels Desktop runs most other x86 operating systems as well. And I haven't met anyone yet who has tested Parallels Desktop's fitness for server virtualization on Xserve.

I have tested, and continue to run in a production setting, two instances of Windows 2003 Server hosted by Parallels Desktop, running on OS X Server 10.4.8 on a 3 GHz Xserve. To skip to the punch line, it works, and it's as fast as all get-out. Parallels does not stretch the truth when it claims near-native performance; Xserve is capable of knocking off any two-socket Netburst (Pentium 4) Xeon server going back at least two years. Compared to Xserve, those Intel boxes eat more electricity and give off more heat than they give back in capacity for work.

Parallels Desktop is a client solution, but I'll tell you how to deal with that. Parallels also carries with it all of the caveats and shortcomings of host/guest virtualization--in which a full OS is required to support virtual machines--but Parallels and Xserve uniquely turn some of those issues into advantages.

Consolidating Windows servers with Xserve is an edifying and productive exercise. I'll tell you how it's going, and exactly how to make it work for your Xserve, or for any Mac that you press into Windows server consolidation duty.

Posted by Tom Yager on December 21, 2006 01:07 PM

December 07, 2006 | Comments: (0)

Leopard Server faces an x86 server market turned inside out

TAGS: Politics, strategy and culture, Servers and storage

I've been glued to the fascinating reality show that is the legal battle pitting SCO and Microsoft against IBM, Novell, and the whole of the server computing universe. There's been an Apple angle all along, but just hacking my way through the dense brush of information has occupied most of my time. I'm finally writing again.

If you haven't been watching, here's my view of the present: SCO Group's UNIX copyright infringement/contract violation litig-a-thon is sinking fast from holes its own lawyers punched in its hull, plus a fusillade of bon voyage torpedoes fired by Novell's rock star legal team. Novell took the offensive against SCO by filing Novell v SCO, and as part of that action, demanded transcripts of correspondence between Microsoft and SCO. Microsoft was an early and enthusiastic endorser of SCO's righteous campaign to protect its intellectual property, and it wielded SCO as a litigious boogeyman to terrify competitors' customers into switching to Windows. Of course I could be reading more into it than the facts support, but see the Steve Ballmer quote at the bottom of this post and tell me what you think. This was par.

Novell set aside one waterborne missile inscribed Novell v Microsoft. While counting down, Novell offered Microsoft one chance at having the firing locks returned to safe position. Novell couldn't have inflicted much lasting damage on Microsoft, but it would have been a public civil trial, a grand humiliating mess, and really, Novell's actions were unpredictable. It's not like Novell didn't have a score to settle, and Microsoft doesn't need to join Intel as a party to root against in an anti-trust case.

Microsoft bought its way out of Novell v Microsoft, I think, by pretty much ceding majority share of the x86 enterprise server market to Novell. The deal, which Microsoft announced as a partnership, is a real ankle-grabber: a) Microsoft will give 70,000 licenses of SuSE Linux Enterprise Server to Windows server customers, b) Microsoft will let Novell ride along on sales calls, c) Microsoft waives its right to sue Novell over intellectual property, and d) well, the D is not something that Microsoft is handing over. But it's an awfully big D.

Novell is going to take back ownership of System V UNIX. It supposedly transferred this to SCO, but a Federal Judge found that the contracts had more holes than paper. Novell has leave to contest the agreements, and if SCO goes belly-up while the matter is in debate, Novell's there to snatch up a very valuable property.

What's so valuable about System V UNIX? For one thing, it's the core of a multi-billion dollar big iron UNIX industry, and every box sold is a license paid to the owner of System V. For another, Linus Torvalds himself said that if an affordable x86 cut of System V existed, there'd be no need for Linux. And in Novell's hands, System V, SuSE Linux Enterprise Server (SLES) and the freedom to lawfully profit from the Microsoft IP and intelligence in Novell's safe, makes Novell the server market's only triple threat. What will Novell do with all of this power? Search me. Nothing that competitors will applaud, unless Novell decides to bring all UNIX and Linux together to take aim at Microsoft. Two things are certain: Novell is going to get a lot bigger, and System V release 4 UNIX will return to x86. Not the puffy Solaris, but the trim, fast, portable, 64-bit, scalable OS that is the commercial server standard. Be unhappy about it if you want to, but in the wild in big production deployments, where you see UNIX, you see System V.

I think that Novell's change of fortune potentially upends the landscape where Apple's server platform is concerned. Linux, with its endless flavors and unbounded bloat and complexity, never hit the big time as an OS preconfigured on new PC servers, a fact that gave Apple an edge. But in Novell's hands and with an exclusive level of Windows interoperability, System V could. I expect that CALs (client access licenses, one required for each device that connects to a Windows server) and overall Windows interoperability will be the battleground in an upcoming Microsoft IP scorched earth campaign against commercial open source, which might not bode well for OS X Server's unlimited client license. Novell bought itself a pass.

Novell could have a clarifying effect on the market. Standing next to carbohydrate-addicted, 32 flavors and then some, back-of-napkin documented Linux, System V release 4 is disciplined, trim, portable, proprietary and no stranger to 64-bitness, zero-tolerance standardization, clustering, virtualization and deployments of massive scale. Linus Torvalds himself said that if there had been an affordable System V UNIX for x86, Linux wouldn't have been necessary. It may prove unnecessary when System V returns to the x86 other than on Solaris.

Don't take me wrong--Apple's competition will change, but Leopard Server is going to do just fine. Leopard Server has nothing to fear from Linux, and I place it ahead of even Windows 2003 Server on the key measures of manageability, cost and openness. Leopard Server may do far better than fine, especially if Xserve evolves at a pace that's on par with the original Xserve. Right now, Xserve's price is fantastic. It's eight-core ready, and Leopard Server's release, if it comes in January as I predict, will hit the market well in advance of Longhorn Server. Apple's server platform is in its strongest position ever.

It's easy for Apple to shoot at Windows, especially in the small deployments that Apple's server platform is going after. But Apple's chief nemesis, not in January but at some unpredictable point in the future, is more likely to be Novell than Microsoft. It's more likely to be some amalgamation of System V UNIX, SuSE Linux and Microsoft technology. System V Windux, if you will.

If Novell strikes you as a fearsome thing, rest assured that Novell has more opportunities to squander its riches as it does to exploit them. The biggest risk that I see is the need to price and position UNIX very carefully to avoid competing with SLES. Portraying UNIX and Linux as complementary will be tough. And Novell will have to start evangelizing enterprise software vendors all over again. The failure to do that led to the downfall of the original x86 System V UNIX. There are so many pitfalls, and Apple has a path through and around them. I expect that Novell's rise will make the market more competitive, forcing Apple to develop unique value in server hardware and software, and that'll be good for everybody.

I leave you with edible the words of Steve Ballmer, taken from a speech given in 2004, a transcript of which is on Microsoft's Press Pass site.

I want to briefly mention intellectual property, not to make this into some kind of big deal, but I do want to emphasize that with all of the dialogue about SCO and Linux and blah, blah, blah, people can get confused about where things are from an intellectual property perspective. When you buy a Microsoft product, we indemnify you from all intellectual property risk, whether it's a patent claim, a copyright claim, legal fees, damages. We say we stand behind our stuff, that's our commitment to you. Sometimes that's a very expensive commitment, if we actually lose the Eolas lawsuit, that will be a $550 million claim for patent violations in the browser. I don't think we'll lose that one but, nonetheless, we're vigorously litigating that. But it's our risk, it's not your risk. We stand behind our software.

If you take a look at where things are in the Linux world, nobody stands behind patent violations for open-source software today, not Novell, not HP, not Red Hat or the other distributors, not IBM. From a copyright perspective, Novell has said they will defend against copyright violations, and they cap all legal damages. HP limits its copyright protection to the SCO claim, same with its indemnification amount. Red Hat offers nothing on copyright, and bounded legal indemnification as it relates to dollar amount. IBM is relatively silent, frankly, on what they're doing on the whole thing.

Why highlight this for you? Let's say you're building a business today, and you know you want to build your business around Windows or Linux, you have to decide what intellectual property risk you want to build, what additional cost might you be pushing to your customers in the future that is unanticipated because nobody stands behind the stuff. We've got very smart partners, companies that I think are very smart, who build devices that have Linux embedded, but they haven't really asked the fundamental question of what IP risk are they taking as the provider of these devices with Linux embedded.

And I'm not just trying to view fear, uncertainty, and doubt. Everybody can go study these issues, but particularly the entrepreneurs starting businesses, like folks in this room, they're worth some study in a way that's probably more important than it would be for a CIO of a large enterprise, because your product in some sense always comes in combination with some underlying piece of system software.

Bless you, Godfather.

Posted by Tom Yager on December 7, 2006 12:35 PM