- Intervention: How to salvage Windows 7
- Desktop Windows: Is it time to "cut and run?"
- It's a bumper crop of VM goodness!
- Hyper-V's Achilles' heel
- How Unix politics are killing OS innovation
- Streaming Office: Death knell for Google Apps?
- Solving the legacy Windows compatibility puzzle
- Windows "Workstation" 2008: One week later
- Windows "Workstation" 2008 Clobbers Vista in Benchmark Testing
- Sun's Innotek acquisition a smart buy
June 10, 2008 | Comments: (0)
Intervention: How to salvage Windows 7
I had an interesting conversation this past week with Ben Rudolph, former marketing guru at Parallels and now a member of the Microsoft Windows 7 team. Apparently, my recent pronouncement of death for his new charge struck a nerve. In addition to setting up a flurry of conference calls with his fellow team members, Ben picked my brain a bit about what I felt they needed to do/say to salvage their next generation desktop OS.
Always willing to interject myself into others’ business, I offered the following "pledge" points for Microsoft to ponder:
- Performance. Pledge that Windows 7 will perform as well as, if not better than, Windows Vista on identical hardware. Show the industry that you can do more than create bloated software. Show us that you can be disciplined when you need to be, that you can write clean, efficient code that doesn't leave us wondering where all the GHz have gone.
You've already stated that Windows 7 will be based on the Vista core. It shouldn't be too hard to squeeze a few more optimizations from that battle-hardened NT kernel of yours while also holding the line on the "cycle creep" that doomed 7's predecessor. Right now, this is Job No. 1 for Microsoft. - Compatibility. Pledge that that Windows 7 will deliver a better migration experience than Windows Vista. After performance, the perceived lack of seamless backwards compatibility is the most pressing issue for IT fence sitters. Encouraging companies to bite the bullet now with Vista because Windows 7 won't help them is not the message you want to be sending.
Leverage some of that slick new legacy application integration technology you've acquired (SoftGrid, Kidaro) to help smooth over the rough spots. There's really no excuse for providing anything less than flawless Win32, COM/ATL/MFC and down level .Net application support, even with UAC enabled. You have all the tools in your portfolio. Time to put them to use. - Usability. Pledge to take all those powerful-yet-buried technologies and bring them to the surface with Windows 7. Start by understanding what's relevant to IT and then focusing your UI development energies accordingly. For example: Previous Versions, a more robust driver model, and deeply-integrated Windows Search? Relevant to IT. Aero (including Flip 3D), Media Center, and the Windows Sidebar? Not so much. Yet which features got the lion's share of PR when Vista launched?
Stop trying to "out-Mac the Mac" on the UI whiz-bang front. Just expose the myriad compelling technologies you already have under the hood and let the IT community be the judge. It's telling when you see the industry media tripping over themselves to lavish praise on something like "Time Machine" when Windows has been doing the same thing better, and with more data storage flexibility, for years.
Finally, Microsoft needs to communicate better, and by this I don't mean just under-promising. The "cone of silence" around Windows 7 has to end. The recent, anticlimactic comments regarding Windows 7's architecture and the subsequent misguided attempts to push Vista at 7's expense have set off a panic in the IT community. Now it's time to open up and start explaining why not reinventing the wheel is a good thing and how, without a kernel overhaul, you can still improve (faster, more compatible, easier to use) on the Vista experience.
It's not too late for Microsoft to salvage Windows 7. As I've stated in the past, Moore's Law is on their side. If they can deliver a richer, more compatible version of Windows Vista, without causing additional code bloat (and maybe wringing some extra throughput out of what's carried over from Vista), they have a decent chance of winning back some of the IT shops they alienated along the way.
Posted by Randall Kennedy on June 10, 2008 03:00 AM
June 03, 2008 | Comments: (0)
Desktop Windows: Is it time to "cut and run?"
Cutting and running. A bad idea for chainsaw massacre movie villains, scissors-wielding grade school students and the occasional foreign incursion. But what about IT? At what point does further investment in a platform or technology cease to make sense?
Until now, I've been advising Vista fence-sitters to wait for Windows 7. However, last week's "big reveal," in which Microsoft finally confessed that Windows 7 will be nothing more than "Vista warmed over," has forced me to reconsider my position. I'm now more convinced than ever that Windows is doomed - at least on the enterprise desktop. What Microsoft's aging (in the U.S., NT is almost old enough to vote) OS needed was a heart transplant. What it got was a new name, a fresh change of clothes and an A.M.A. discharge from the ER.
It's not enough. The IT community deserves better. Microsoft's loyal customers deserve better. The folks from Redmond owe us a fresh start; a clean break; a path to the future free from legacy baggage. Basically, they owe us a new Windows.
"MinWin" was a good start. Unfortunately, Microsoft executives seem to lack the will to see it through. It's easier to keep re-warming the same stale bits over and over again, safe in the knowledge that the Windows customer base will lap it up all the same. Monopolization has its privileges.
Of course, the secret to maintaining a de facto monopoly is to not push your customer base too far. Every industry segment has its breaking point, a fact Microsoft learned the hard way with Windows Vista. With Vista, they ignored IT, choosing to instead kowtow to big media and the DRM crowd. The result was an unprecedented backlash as angry IT shops spurned Vista and drove the Save XP campaign to international attention.
Now we learn that Microsoft's one chance to get it right - to atone for its sins and perhaps salvage some modicum of respectability - is really just another PR stunt. Windows 7 will be Windows Vista "Part Deux." Same clumsy, monolithic architecture. Same crippling legacy baggage. Whatever hope may have been sparked by the whole "MinWin" fantasy has now been officially snuffed out.
If I sound frustrated it's because I am: At Microsoft for refusing to leverage its vast technical resources to address Windows' myriad shortcomings; and at the customer base for agreeing to take each new batch of Windows lemons and make lemonade.
When will the Windows community wake up and realize how much power it has? We've already demonstrated what can be done when enough of us decide that, hey, we really don't like lemonade all that much. Vista is a debacle precisely because we decided "enough is enough" - and because a viable alternative (XP) still existed. Microsoft's response? Try to squash the "rebellion" by pulling the rug out from under us (i.e. XP's impending expiration date) . Talk about sour grapes!
That's why I say it's time for the Windows community to take a hard look at alternative platforms, like Linux and Mac OS X. It's over there, on the other side of the fence, that the real innovation is occurring. By contrast, Windows - including the over-hyped version 7 - is an architectural dead end. We, as a community, need to accept this fact and move on.
Posted by Randall Kennedy on June 3, 2008 03:00 AM
May 20, 2008 | Comments: (0)
It's a bumper crop of VM goodness!
Life in the blogosphere has its ups and downs. Some weeks there's little to write about. Others, you're inundated with too many juicy story ideas to count.
Last week was all about virtualization. After savaging Microsoft's Hyper-V Achilles' Heel, I was pleased to discover a new beta release of VMware Workstation 6.5. Also dropping in to pay a visit: Virtual PC 2007 Service Pack 1. Truly a bumper crop of leads to follow-up!
First, I just got done putting the latest VMware 6.5 pre-release (Beta 2?) through its paces. The earlier beta was already pretty solid (see my preview for the Test Center), so I wasn't expecting all that much in the way of improvement.
One thing I did want to test again: 3D video hardware acceleration. In the earlier beta I was able to fiddle with a few DirectX tests -- including the Direct3D 7, 8 and 9 cube thingy in dxdiag -- however, the experience was quite buggy. The VM would often freeze or crash, and I never made it through the entire test suite without some sort of bug or issue.
Clearly, VMware was listening. The new build (91185) is rock solid, allowing me to complete the full range of Direct3D tests as well as play some legacy games (Quake II, Starfleet Command III) under Windows XP Professional (SP3). Otherwise, the basic feature set remains the same. There's still the slick new Easy Install option, which worked great with both XP and Vista. However, I'm not yet sold on the whole Unity business. The amount of shearing and other visual weirdness I encounter when enabling Unity makes the feature all but unusable for me - that, and the fact that VMware crashes whenever I try to enable it while CubeDesktop is running.
Oh well, at least VMware is making progress on its 3D acceleration feature. Virtual PC SP1, on the other hand, is really an update in name only. Aside from an expanded set of supported host OS (Windows Vista SP1, XP SP3) and guest OS (Windows Server 2008 Standard Edition, Vista SP1, XP SP3), the only real change is to the Virtual Machine Additions, which are now at version level 13.820 (up from 13.803 with the original Virtual PC 2007 release). Otherwise, the product remains effectively unchanged, which also means that it sill doesn't support 64-bit guest OS and only pays lip service to Linux and other non-Windows platforms.
The other major contender in this space -- VirtualBox -- seems to be benefiting from its recent acquisition by Sun. Version 1.6 was released at the end of April and brings the full force of Sun's engineering prowess to this previously obscure VM solution. One feature, in particular, that seems to have received some polish is Seamless Windows. Much like Unity under VMware, Seamless Windows lets you run applications from a VM directly on the host desktop. The only difference is that Sun's implementation actually works well. There's none of the window shearing that makes Unity so difficult to stomach and performance seems at or near that of a locally executing application. And most importantly, it doesn't crash when enabled with CubeDesktop running.
Overall, VirtualBox 1.6 is a solid release and one that shows how serious Sun is about producing a competitive, Open Source alternative to VMware Workstation. Check it out.
Posted by Randall Kennedy on May 20, 2008 03:00 AM
May 13, 2008 | Comments: (0)
Update: Since first posting this entry I've been in communication with Microsoft about the BSOD I encountered with the ATI installer. In an effort to assist them in their investigation, I recreated the scenario that caused the failure and proceeded to capture memory dump information for them to dissect. It's also worth noting that bug continues to be reproducible under Hyper-V Release Candidate 1. It seems to have something to do with the installer attempting to probe the hardware in order to identify the specific device make, model and revision - a common task performed by driver installation programs from myriad vendors. My advice is that, if you must deploy Hyper-V, make sure you perform any driver updates - including those for plug & play or hot-pluggable devices - with the server in an idle state and with any hosted VMs shut down or with their running state saved to disk. Given the catastrophic nature of this failure mode, and the general fragility of the Hyper-V driver architecture, it is my opinion that you simply cannot be too careful when maintaining the host/root partition of a Hyper-V server
A house of cards -- that's how I'd describe the current state of the Windows device driver ecosystem. With so many Windows-compatible devices and so few competent driver developers, it's no surprise that hunting for driver updates has become a necessary part of every power user's skill set. Most of the time, the search ends in frustration: Either the new driver doesn't correct the existing problem(s) or, worse, creates a set of new ones. And now, with the introduction of Hyper-V, we have a whole new failure vector to think about.
In a nutshell, one of Hyper-V's advertised strengths -- the host partition's ability to work with generic Windows device drivers -- is also its greatest weakness. That's because the quality level of Windows device drivers, especially those from third-party developers, is notoriously inconsistent.
I found this out the hard way while experimenting with the Hyper-V Release Candidate on a newly configured Windows "Workstation" 2008 system. After enabling the Hyper-V role in Server Manager, I made the mistake of trying to install the latest ATI Catalyst (8.4) software for the system's X1300 display adapter. The resulting Blue Screen of Death was both alarming (I hadn't seen one of these in months) and puzzling: I had successfully installed this driver before, on the same system, without incident. The only difference this time around was Hyper-V (uninstalling the role and rebooting allowed me to complete the driver installation).
Even more disturbing was the fact that I had just finished watching an old (December 2007) Channel9.com interview with Mark Russinovich, a Technical Fellow at Microsoft and one of the smartest guys I know. In the interview, Mark talks about Hyper-V and how its ability to leverage existing Windows drivers in the host partition gives Microsoft a competitive advantage over certain unnamed competitors (read: VMware), which require custom drivers for their proprietary hypervisor OS layer.
It all sounds great on paper, until you realize that it effectively places Hyper-V -- and the rest of Microsoft's virtualization architecture, for that matter -- at the mercy of the single most glaring weakness of the Windows ecosystem: third-party device driver developers, most of whom have no idea what Hyper-V is or how to avoid tripping over it during driver configuration/installation.
I point this out because it runs counter to everything that makes VMware's ESX platform so compelling. With ESX, you get, effectively, a black box: a proprietary environment, but one with its own, rigorous testing and development model. The pieces that go into that box -- the drivers and services that extend the Console OS layer (which is, itself, a derivative of Linux) -- are carefully vetted to ensure at least a baseline level of robustness.
By contrast, the Windows device driver landscape is more akin to a Wild West shoot-out. And while you can try to minimize the risk by sticking to Windows Hardware Quality Lab (WHQL)-certified products, there's no guarantee that they'll work reliably under the added stress introduced by the shared VM bus architecture on which Hyper-V is built. Eventually, something is going to cause a conflict, resulting in the kind of catastrophic system failure I experienced during the aforementioned ATI driver installation.
Bottom line: What Microsoft needs is more and better certification options. The company needs to expand WHQL to include Hyper-V testing and/or create a parallel program that further tests WHQL candidates for Hyper-V compatibility. Until then, it'll be hard to take its virtualization plans -- desktop or server -- seriously.
Posted by Randall Kennedy on May 13, 2008 03:00 AM
May 07, 2008 | Comments: (0)
How Unix politics are killing OS innovation
It's frustrating. Each time I think I've found a viable long-term replacement for my Windows development and testing environment, along comes some stumbling block to trip me up. And while these hurdles are mostly technical, the "root" causes (pun intended) are almost always political in nature.
Case in point: My various bouts with the disease known as "Ubuntu-itis." Every six months or so I get the urge to jump ship and join with the great hippie masses swaying to the open source beat. But then "the man" has to come and spoil the fun. All that finger pointing about the ACPI bug -- nVidia pointing to the kernel team pointing back at nVidia, with me pointing at Canonical for failing to sort the whole mess out. It's like trying to decipher that Battlestar Galactica "Last Supper" picture: Everyone's got a knife out for someone else (FYI, I'm leaning towards the doe-eyed Dualla as the final Cylon).
This week it's OpenSolaris, the new kid on the FOSS block that may not be quite as "open" as Sun would like us to believe. That's because, despite public statements to the contrary, Sun may not "own" the Unix IP in the way that most of us think of ownership. In fact, it's looking more and more like the only true IP holder is Novell, the same company that just got done grinding SCO's bones to make its bread. If they get hungry, might they turn on Sun? How much damage could Novell inflict if Sun decides to go ahead with a GPL-3 license for OpenSolaris and the folks from Provo object? More importantly, would you feel comfortable adopting an OS platform that may get litigated out of existence before its first birthday?
It's a shame because I actually like OpenSolaris. The 2008.5 release is polished, easy to install/configure and seems chock full of promise. During my own preliminary testing, the Gnome-based UI had me feeling right at home, though re-inventing the package manager wheel seemed a waste. Most of my devices were supported (I tested on a Dell OptiPlex 745 instead of my usual XPS notebook), including such esoteric stuff as the bay-mounted 12-in-1 memory card reader (Memory Stick Pro Duo support -- yeah!). Even VirtualBox seemed to work correctly, which isn't surprising since Sun now owns the code base.
Overall, it was far more pleasant experience than my various PC-BSD misadventures. I got the sense that Sun actually knows what they're doing with Unix, which makes the thought of Novell stomping this nascent Linux challenger into oblivion that much more disturbing. Here's hoping that OpenSolaris survives and serves to keep the increasingly insular Linux community -- including the "hear no evil, see no evil" crowd that controls the kernel -- honest.
Posted by Randall Kennedy on May 7, 2008 12:10 AM
May 01, 2008 | Comments: (0)
Streaming Office: Death knell for Google Apps?
It's the holy grail of punditry: You make a bold, unconventional prediction that's way out in front of an emerging trend, then get to gloat as that prediction comes true. In my case, the prediction involved Microsoft's recently (at the time) acquired SoftGrid technology. I saw the acquisition as a clear sign that Microsoft was gearing up to deliver a subscriptions-based, hosted licensing model for Microsoft Office.
Of course, not everyone agreed with me. In fact, the CTO of a key Microsoft competitor all but dismissed the possibility on technical grounds. SoftGrid, he said, was an "inside the firewall" technology, and Microsoft would be hard pressed to make it work over the web.
Fast forward to today and we now hear that Microsoft intends to do exactly what the "nay-sayers" said they wouldn't: Use its newly re-branded Microsoft Application Virtualization (MAV) technology to deliver Office and other traditional "fat client" applications as subscriptions-based services, with the Internet/web (along with some of those massive data centers the company has been building) providing the distribution infrastructure.
Now it's time for another bold prediction: Streaming Office will clobber Google Apps, eventually driving the search giant out of the hosted applications business altogether.
Consider the following advantages to a Streaming Office suite:
- Full Office Functionality - MAV encapsulates the entire sequenced application. This isn't some "web-based" Office knock-off. It's the real deal: Microsoft Office in all it's sophisticated, class-leading, standard setting (flaunting?), enterprise desktop-dominating glory.
- True Offline Operation - Though Streaming Office may initially come to you over the wire, the actual application bits get cached locally by the MAV agent. This is the same MAV agent that hosts the virtualized application runtime, so offline operation is entirely seamless. As any veteran SoftGrid user can attest, it's as good as - if not better than - having Office installed locally.
- It's the Future of Windows Application Deployment - Everyone knows that Microsoft needs to "fix" the Windows runtime environment. All of that legacy application baggage is starting to drag down the client, making each new Windows version "fatter" and more kludge-like. I believe that MAV is one solution Microsoft may be mulling over as a means to preserve backwards compatibility while they tighten the native runtime so that only well-behaved code (including the MAV runtime) can execute.
With this kind of R&D investment you can be sure that Streaming Office - which is a byproduct of MAV's continued evolution - will be a high quality product and receive tremendous, cross-divisional support (just like Office does today).
Faced with a decision between a watered-down, limited, web-dependent pseudo-suite and the full power and richness of Microsoft Office, users will flock to the Microsoft camp - especially if the company prices a "pay as you go" Office aggressively. And then it will be "game over" for Google Apps and its ilk.
Just remember: You read it here first!
Posted by Randall Kennedy on May 1, 2008 11:23 AM
April 15, 2008 | Comments: (0)
Solving the legacy Windows compatibility puzzle
There's been a lot of chatter lately about how Microsoft needs to start over with Windows. Many point to the (NT) code base's 16-year history and how the need to maintain backward compatibility is hampering efforts to move the platform forward. According to these critics, a clean break is necessary in order to stop the kind of bloatware madness that so crippled Windows Vista. Dump the creaking legacy that is the Win32 API/ATL/MFC, they say, and solve the compatibility riddle through VM technology.
While I can appreciate the logic behind these assertions, I don't agree with the proposed remedy. For starters, the inferred replacement for these outmoded APIs is Microsoft .Net. As those who are familiar with .Net programming will attest, so-called managed code runs like a slug in a molasses bath -- a least on the client side of the fence (I've personally had success with .Net for server application development). Convincing developers to dump tried and true -- if somewhat anachronistic -- programming models in favor of a fatter, slower runtime is even harder than it sounds.
Then there is the issue of application fidelity. Despite advancements in VM technology -- most notably, support for accelerated 3-D graphics in VMware Workstation 6.5 -- the fact remains that running legacy applications in a virtual machine is far from ideal. In addition to the sizable overhead (several hundred megabytes of RAM to house the applications and their supporting OS images), you lose the seamless integration of a natively executing Windows program. Popular workflow functionality, such as COM/OLE automation, becomes nearly impossible to implement across VM boundaries. Even simple tasks, such as dragging and dropping a data file onto an application window, take on new levels of complexity. No matter how cleverly disguised, these virtual walls of isolation will simply aggravate users who are accustomed to a consistent operational experience.
It would seem that Microsoft is faced with a dilemma. On the one hand, it's being called upon to move the Windows platform forward by eliminating the compatibility baggage, while on the other, it's stuck with the expectation that Windows will somehow continue supporting legacy applications, most likely through virtual machine technology. But is VM compatibility -- with its inherent shortcomings -- the only option?
Fortunately, the answer is no. There is another compatibility avenue that Microsoft might consider: application virtualization. Instead of building in a complex VM compatibility box, Microsoft could simply tweak its SoftGrid-derived Microsoft Application Virtualization (MAV) technology to provide a controlled runtime environment in which to execute legacy applications.
Here's how such a solution might work:
- Microsoft would integrate the MAV client and Application Sequencer tool with the base Windows OS image as a compatibility subsystem -- much like it provided OS/2 and POSIX support in the past.
- When a legacy application installation request is detected, Windows would fire up the Sequencer subsystem and capture the installation to an MAV image. The process could be further buttressed through the inclusion of an extensible compatibility library of Sequencer tweaks and so forth.
- The resulting application will still run in a virtualized state, but without the overhead and hard boundaries of a traditional virtual machine.
Such an application could respond to shell events (drag and drop, context menu selections), interact with other applications via COM/OLE, and generally preserve the fidelity of a native application -- all without mucking up the file system and Registry hives or otherwise creating the kind of conflicts/security holes that the "clean break" advocates like to squawk about.
It's a best-of-both-worlds scenario, one that would allow Microsoft to isolate troublesome legacy applications (or those from ISVs that still have not abandoned the old model) and actively evolve the native Windows runtime without concern for breaking the legacy application base or saddling users with an imperfect VM-based solution. The company has all the pieces. Let' see if it's smart enough to put them together in time to salvage the Windows platform.
Posted by Randall Kennedy on April 15, 2008 03:00 AM
March 10, 2008 | Comments: (0)
Windows "Workstation" 2008: One week later
It's been just over a week since I took the plunge and nuked my Vista SP1 installation in favor of a desktop implementation of Windows Server 2008 (aka Windows "Workstation" 2008). So far, it's been a smooth ride. All of my core applications are working flawlessly, including SQL Server 2005, Visual Studio 2008 and Office 2007. In fact, outside of the multimedia issues I mentioned in my previous entry, I'd be hard pressed to find fault with the experience.
Some observations:
- Visual Studio 2008 flies on "Workstation" 2008. This is no joke. I was ready to throw in the towel on Studio 2008 under Vista. Now, time-consuming operations -- like selecting a complex ASP.Net object on a heavily populated Web form -- that would bog down under Vista just snap right along under 2008. The IDE loads faster, as do my projects. As far as Studio 2008 is concerned, "Workstation" 2008 has delivered an all-around speed boost and has proven to be a real productivity enhancer.
- VMware Workstation also runs better on 2008. I'm a big VM user because I need to test my code against so many different OS permutations. With "Workstation" 2008, juggling even large, multi-gigabyte VMs is a breeze. You immediately sense the performance improvement as even hard-to-virtualize OSes, like Vista x64, run more smoothly. It's made the process of testing my latest revisions that much more pleasant.
- The OS never feels "sluggish." Under Vista, you inevitably reach a point where the OS starts to "slow down." It may be after a day of heavy use or even a week of on-again/off-again (or in the case of my Dell XPS M1710 "notebrick," suspend-again) computing. But eventually you'll feel the need to reboot and start clean. With "Workstation" 2008, I've yet to encounter such a malaise. Whether it's better memory management or simply a more mature, polished code base, "Workstation" 2008 delivers a level of robustness and consistency that Vista can't touch.
Of course, the really bizarre aspect to my "Workstation" 2008 experience is that, from a technical standpoint, it doesn't make sense - or at least, it shouldn't. With the same kernel (as of Vista SP1), the OSes should in fact behave similarly.
To be sure, the Windows Server 2008 kernel is configured differently at boot-up. Different switches are applied to tune the kernel for server duty. In fact, this was the norm with all "NT" client and server releases prior to Windows XP: A shared code base differentiated primarily by boot-time tuning of the kernel image. However, it's hard to fathom how tuning alone could cause what is essentially the same OS to behave so differently in identical workload scenarios.
Something else is going on here - or perhaps isn't going on. Some of my "Workstation" 2008 compatriots have speculated that DRM is to blame, that Windows Server 2008 boots with less of the DRM plumbing than the consumer-oriented Vista and that this, in turn, frees up cycles for more important tasks (like the applications you're running to actually get some work done).
Whatever the cause, I will continue to dig into the differences between the Vista and "Workstation" 2008 runtime environments. In particular, I'll be looking for evidence of OS layers that might be missing from the overall "Workstation" 2008 stack, layers that might be the source of the up to 17 percent performance hit that tests show Vista's particular flavor of the "NT 6.0.6001" kernel introduces.
I'd also like to hear what other users are finding as they experiment with Windows "Workstation" 2008. Let's get the dialog going and see if the initial hype stands up to some old fashioned scrutiny. Fire away!
Posted by Randall Kennedy on March 10, 2008 12:21 PM
March 05, 2008 | Comments: (0)
Windows "Workstation" 2008 Clobbers Vista in Benchmark Testing
Just when you think you've got it all figured out, along comes another curve ball to keep you guessing. This time around it's a set of new benchmark test results from our old friends at the exo.performance.network. Apparently, these guys heard about the brouhaha surrounding some Microsoft's engineer's blog post about using Server 2008 as a "Super Workstation OS" and decided to put the concept to the test. They loaded each OS onto identical hardware, tweaking the Server implementation to make it look and act like Vista. Then they tested them under a variety of productivity and multitasking scenarios.
The net result: Windows "Workstation" 2008 (i.e the tweaked Server 2008 installation) trashed Windows Vista with Service Pack 1, outscoring its desktop sibling by as much as 17%.
You can read all the gory details over at the exo.blog site. Suffice to say that there's a new option for power users who are fed-up with Vista's sluggishness: Windows Server 2008, or as I'm now calling it, "Windows NT 6.1 Workstation."
I made the switch myself this past weekend in support of my upcoming feature on the "Top 10 Reasons (Not) to Switch to Vista." And after just a few days running this beast I can honestly say that I will never go back to the sad piece of "bloatware" that is Windows Vista Ultimate. "NT 6.1 Workstation" is simply a superior OS. It's faster, feels more responsive and seems to do a much better job of managing the 4GB of memory I've thrown at it (Vista always seemed a bit page file-happy, even with the 4GB).
To be sure, there are things I miss about Vista. For starters, there's no Media Center. I had to scrounge around for a bunch of 3rd party CODECS just to play a DVD. Nor are any of the Vista Games available on Server 2008 (Solitaire, I'll miss you!). Useful Vista accessories, like Fax & Scan and the Windows Sidebar, are MIA, while some 3rd party tools are just plain broken (e.g. I had to downgrade my Skype install to version 3.2 since anything newer would crash consistently).
On the flip side, I can now run any and all Microsoft server applications locally, which makes debugging things like ADSI and SQL Server's Integration Services that much easier. In fact, as a pure development platform for Windows applications, Server 2008 is without equal. There's the performance (this OS really does rock). And, of course, you can tap Hyper-V to create a near-bare-metal VM environment. Visual Studio 2008 flies on this release. There really is no downside, especially if you have an MSDN account or are similarly licensed so you can use Windows Server 2008 on your own box for development purposes.
Given all the positives, I can't imaging why Microsoft didn't break-out the consumer aspects of Vista and offer a leaner, meaner "Workstation 2008" version for us hard core types. It would have gone a long way towards satiating our thirst for a "meatier" desktop Windows while rendering much of the "Save XP" messaging moot.
Maybe Microsoft just got lazy...oh well, at least now there's an alternative for those of us who are fed-up with Vista but who just can't live without Aero and the rest of the "6.x" Windows GUI.
Posted by Randall Kennedy on March 5, 2008 10:35 AM
February 12, 2008 | Comments: (0)
Sun's Innotek acquisition a smart buy
It was bound to happen eventually. Somebody, somewhere, was going to see the light and snatch-up the last of the original virtualization pioneers, Innotek, makers of the VirtualBox Virtual Machine platform for Windows, Linux and Macintosh systems.
That somebody is Sun Microsystems. And based on the wording of their press release, I'm convinced they did so for all the right reasons.
For starters, they recognize that Innotek's strength is in the desktop space. Though the German virtualization developer has been working on some more server-centric models (including at least one 32-core implementation), they're greatest success has been as a low or no-cost alternative to VMware Workstation and Microsoft Virtual PC.
But perhaps even more important is Sun's recognition of Innotek's commitment to developers. VirtualBox has long been the preferred solution for open-source programmers seeking to "roll their own" virtualization platforms. The combination of componentization (virtually every layer of VirtualBox can be scripted or accessed programmatically) and GNU public licensing of the VirtualBox source code makes for an attractive package that Sun can use to further its own FOSS goals.
So, congratulations (I seem to be saying that a lot lately) to Innotek, and a hearty "bravo!" to Sun for seeing the value in VirtualBox and its potential as a building block for a robust Desktop virtualization strategy.
Posted by Randall Kennedy on February 12, 2008 11:36 AM
February 08, 2008 | Comments: (0)
Microsoft's Caligari acquisition triggers flashback
When I heard the news that Microsoft had acquired Caligari Software, makers of the trueSpace 3D modeling application, I thought for a moment that I had slipped through some rift in space/time. It was once again 1996, and I was fresh off a stint with IBM's Personal Software Marketing division. My new wife was pregnant with our first child and we were working together on a side project involving VRML and Biochemistry.
The project consisted of developing 3D VRML models of various building-block molecules. For those of you unfamiliar with the acronym, VRML stands for Virtual Reality Modeling Language, a kind of 3D file format for rendering objects or landscapes over the web. In practice, you'd use a browser plug-in to parse the file and render the "world" in an interactive (i.e. you could navigate within the 3D space) window inside a web page. VRML even had its own funky URL syntax that would trigger Internet Explorer or (at the time) Netscape Navigator to load the plug-in.
It was new. It was cool. And it was entirely impractical.
For starters, VRML files tended to be fat, much fatter than your typical HTML document. In fact, the files were *so* large that the plug-in makers all adopted a compression scheme involving gzip in and effort to lower the network overhead. However, even with compression (and other bandwidth-saving tweaks, like loading only portions of the VRML object based on view "distance" and the level of detail required), VRML still never caught on outside of a few "gee whiz" 3D demos. We ended-up shelving the project after we discovered another company had already released a competing library using a more powerful (it was designed specifically to render molecular models and had features that that VRML couldn't touch), proprietary -- yet freely downloadable -- plug-in.
I bring all this up because, at time of the aforementioned project, my tool of choice for VRML modeling was trueSpace2 (and its "freebie," VRML-only sibling called "Fountain" -- trueSpace2 didn't support VRML authoring directly). Many long nights were spent wrestling with Caligari's oddball UI decisions (they basically threw much of the Microsoft application design guide out the window and rolled their own). It was an interesting time in my life, and my involvement with VRML ultimately led me to write a book about the technology: "Instant VRML Worlds" (look it up).
Returning to the present for the moment, it's my understanding that Microsoft has acquired Caligari to help with its Virtual Earth project. I'm guessing they want to tap Caligari's expertise in LOD (Level of Detail) management and other VRML-like technologies that will be critical to the success of the their upcoming 3D version of Virtual Earth.
Congratulations to Caligari for getting picked-up by the Redmond behemoth. It's always nice to see an "old friend" make it big ...
Posted by Randall Kennedy on February 8, 2008 08:37 AM
February 02, 2008 | Comments: (0)
Dell servers + virtualization: a disastrous combination
OFF TOPIC RANT MODE: ON
I like Dell. Over the years I've owned several dozen Dell systems, including notebooks, desktops and servers. So naturally, when it came time to spec out a new server for the exo.performance.network Web site, I went with what I know and bought a shiny new PowerEdge 2950.
The 2950 is a sweet box. Our unit is tricked-out with 8x 2.66GHz 45nm Xeon cores, 8GB of RAM and 1.2TB of usable disk space in a six-disk RAID configuration. It's powerful and also deceptively sleek - the entire unit fits in just 2U of rack space, saving us on co-location costs.
Unfortunately, it's also useless for any kind of server consolidation/virtualization tasks. That's because the unit ships with a pair of integrated Broadcom NetExtreme II gigabit Ethernet NICs, the most virtualization-unfriendly NICs on the planet.
Our original plan was to run SQL Server on the bare iron using Windows Server 2003 R2 x64 edition while hosting our Web site (which includes some proprietary 32-bit ASP.NET components) in a Windows Virtual Server 2005 R2 SP1 virtual machine running the 32-bit version of said OS.
However, after a week of struggling to get the NetExtreme II adapters to maintain the Web site's connectivity (the virtual machine would randomly drop its connection and go "deaf" every few hours, requiring a VM reboot), we had to scrap our plans and instead run the Web site code directly on the Host OS (a process that involved reconfiguring the 64-bit IIS service to run32-bit worker processes - i.e. UGLY!).
In the end, we got the site up and running sans virtualization (see for yourself at www.xpnet.com), but I've now been soured on Dell as a server vendor. Not only was their technical support entirely unhelpful (basically, they passed the buck to Broadcom, even though the adapter is integrated with their hardware), we could find no mention of the problem in any Dell or Broadcom knowledge base entries or support articles. Yet a search on Google yields copious examples of customers encountering the very same problem with this, and similar, Dell servers.
Bottom Line: If you're looking to consolidate servers through virtualization, buy HP..or IBM...or Sun. Anyone but Dell. And stay away from Broadcom NICs! You'll thank me for it...
OFF TOPIC RANT MODE: OFF
Posted by Randall Kennedy on February 2, 2008 11:01 AM
January 22, 2008 | Comments: (0)
Virtualized MS Office: just around the corner?
I found the following tidbit buried at the bottom of a recent Microsoft Press Release. Titled "Microsoft Announces Vision and Strategy to Accelerate Virtualization Adoption," the release goes on to note:
Microsoft announced today that the 2003 and 2007 versions of the Microsoft Office system are supported when running in both Microsoft Application Virtualization 4.5 and SoftGrid Application Virtualization 4.2.
This is interesting in that it signals a major shift in Microsoft's view of its nascent application virtualization platform. No longer a "fringe" technology, "MAV" is now an accepted part of the Windows ecosystem and a key building block in the company's overall virtualization strategy.
It also sends a clear message to IT shops: It's now OK to start building around the MAV platform. To date, most shops have been operating under the "old rules," essentially using MAV's underlying SoftGrid technology to deliver applications as if it were still a third-party solution. Now, with Microsoft's formal endorsement of the technology as a delivery vehicle for its most precious of cash cows, Microsoft Office, the training wheels are finally ready to come off. The MAV way is now officially co-equal to the traditional, "fat" client model that has defined Office installations for nearly two decades.
Of course, the story doesn't end there. Declaring parity between MAV and non-MAV Office deployments is just the first step in a much broader strategy designed to convert Microsoft's entire business line to a more predictable (from a revenue projection standpoint), manageable subscriptions-based delivery model. That's why Microsoft is working so hard to make sure IT shops get comfortable with application virtualization. In the not-so-distant future (read: 12 to 18 months, or about the time that "Office 13" ships), MAV will overtake the classic "local" install to become the preferred means for deploying Microsoft's latest and greatest.
In the meantime, Microsoft keeps seeding the market with bits and pieces of the SoftGrid carcass. First came the Microsoft Desktop Optimization Pack (MDOP), essentially a loss leader to hook its big-name Software Assurance accounts. Then this past week, it released its MSI Packager tool that allows IT shops to decouple MAV-sequenced applications from the complex back-end streaming components that made SoftGrid so hard to deploy.
I expect the full-court press to continue throughout the coming year, with Microsoft "gifting" more and more of MAV until its biggest customers find the bait irresistible and start dabbling with the technology. Things to watch for: A "free" version of the MAV client software (capable of running MAV-sequenced applications in offline mode); trial versions of key Microsoft products becoming available as MAV-sequenced packages; and the beginning of the real endgame when it's revealed that Microsoft Office 13 will be available preconfigured for MAV deployment.
And remember: You heard it here first!
Posted by Randall Kennedy on January 22, 2008 12:14 PM
January 15, 2008 | Comments: (0)
Thinstall: the "hairball" that chokes VMware?
I guess it was just a matter of time. When I last reviewed Thinstall, I noted that, while the technology was useful for certain niche scenarios, what the company really needed was to be swallowed-up by a larger player. Now that VMware has stepped up and filled that void, it's time to assess just what it is that the author of the ill-fated "Virtual Desktop Initiative" (VDI) has ingested.
Some might see the acquisition as a "tasty morsel," a way for VMware to expand into the nascent Application Virtualization space by purchasing a smaller (20+ people at last count) player with an outsized presence in the market. I, on the other hand, see a potential "hairball" in the making.
Thinstall's self-contained virtualization layer is a version control nightmare, one that has hamstrung the company's ability to penetrate the broader enterprise computing market. This is an architectural problem: The entire Thinstall runtime -- along with any applicable Access Control Lists (ACL) -- is embedded within the packaged executable. Changing any part of the package (patching the code, modifying the ACL) requires that you regenerate it from scratch. From an IT perspective, it's an inflexible mess.
Of course, VMware is probably willing to overlook these caveats if doing so allows it to continue its endless quest to "topple Microsoft at all costs." After all, the Redmond gorilla did its own gobbling a scan 18 months ago, scooping up the problematic SoftGrid in an effort to refashion its own delivery architecture. And in the battle for roadmap bullet points and RFQ checkboxes -- where missing a tick or two might mean the difference between a sale and and empty stomach -- maybe it's better to cough-up the occasional "hairball" and hope nobody notices the stain on the carpet.
Posted by Randall Kennedy on January 15, 2008 08:55 AM
January 09, 2008 | Comments: (0)
One step closer to a SoftGrid future
The release of Microsoft's MSI packaging tool for SoftGrid-sequenced applications signals a major step forward in the company's quest to establish its application virtualization technology as a new de facto delivery vehicle for Windows applications. By making both the MSI utility and the underlying SoftGrid client and sequencer applications available via its Web site, Microsoft is effectively seeding the enterprise market with its own brand of application virtualization.
Note: You may wish to check with Microsoft regarding any licensing requirements or restrictions associated with the above downloads.
The net result should be a groundswell of interest in SoftGrid as IT shops check out the newly decoupled (from its server back-end) virtualization engine. With the MSI utility, organizations can deploy SoftGrid-sequenced applications like they do any other MSI-packaged application: Via SMS/ESD, network share, "sneaker-net" (CD/USB Key), etc. It's an aggressive move, one that could easily be interpreted as a latent "shot across the bow" of virtualization competitors Thinstall and Symantec.
I tested the new MSI utility under Windows XP, using the client and sequencer packages from the above links. After sequencing Adobe Acrobat Reader 8.1 I was left with an ~80MB compressed "SFT" file which, along with the corresponding project file, were then scanned by the MSI utility as it created a compatible Microsoft Installer package. I then copied the new package and the SFT file to the target system and installed Acrobat Reader using the MSI package directly. Once installed, Acrobat loaded and behaved like any other SoftGrid-sequenced application.
Some caveats:
- To install an application from a package created by the MSI utility you need to have the latest (4.2.1) SoftGrid Client installed on the target system. See the client link above for more info.
- When installing the client you'll need to enable "Stand Alone Mode" by setting the "MSIDEPLOYMENT=TRUE" parameter on the SoftGrid client installer package's command line. See the MSI Utility Admin guide for command line examples
Overall, the new SoftGrid MSI utility works as advertised: It allows you to take a SoftGrid-sequenced application and distribute it without the complex back-end streaming and access control infrastructure normally associated with a SoftGrid deployment. As such it effectively decouples the client from the server and creates a potent, self-contained delivery vehicle for Microsoft's brand of application virtualization.
Posted by Randall Kennedy on January 9, 2008 04:48 PM
December 29, 2007 | Comments: (0)
Everyone's got an end of year list, so...
- WGA Nightmares
After nearly losing my primary development workstation to this Orwellian scam I began to appreciate why so many power users have switched to Linux. Hopefully, Microsoft will honor its commitment to remove the Vista "kill switch" with SP1.
- Linux ACPI Woes
I was *this* close to switching to Ubuntu. Then I realized that the power management scheme had more in common with NT 4.0 (i.e. effectively non-functional) than NT 6.0 (i.e. a very functional Vista). Perhaps they'll get it right with "Hardy."
- Broken Vista Backup
Inexcusable. If something's stored in a user's personal *data* folder structure (i.e. Document, Picture, Music, etc.) it should be backed-up. Period. That I have to ZIP-up my work each day (including all my my .aspx files) to ensure that Vista's Backup actually picks-up the data is ridiculous. What could've been, what might've been...
- VMware Arrogance
Note to EMC: Rein these guys in a bit, will ya? All that talk about VDI and world domination is getting a bit stale. The real threat to Microsoft's hegemony will come from the browser, not a virtual machine.
- Longhorn Nostalgia
Let's see: Take some aging bits from an incomplete (and now defunct) Windows beta release, mix-in some clever photo-chopping and a few install script tweaks, and voila! You have a hunk of steaming, semi-functional crap that is *still* subject to international copyright laws. Of all the stupid schemes...
- User Account Control
Everyone hates UAC for the wrong reasons. Yes, its annoying. And yes, you're worried that users will just blindly click-through the warning dialogs. However, the real issue is the false sense of security: Even if you pay attention to every elevation prompt you're still not secure. It took less than six months for some clever hackers to discover a potential loophole in UAC's "non-elevated" administrator scheme. Take a cue from the Linux crowd: Dump Vista's default security scheme and run as a true "Standard User" whenever and wherever possible.
- Forking Linux!
Talk about your hyper-sensitive types! I had the audacity to comment on the internal politics of the Linux community! Shame on me! On the plus side, the resulting firestorm permitted me to achieve that rarest of journalistic trifectas: The 3-column "Dvorakian" bait and switch. Thanks, guys!
- 64-bits Gets the Shaft!
First it's the lousy driver support. Then it's being shut-out of various "Live" beta programs. However, the final straw was learning that Microsoft would not be supporting Vista x64 as a SoftGrid client platform. How can Microsoft expect us to take its 64-bit offerings seriously when they keep giving users of Vista x64 (of which there is a growing base) the shaft?
- Vista SP1
Once the cure-all that would save Vista, Service Pack 1 is turning out to be a major non-event. Performance improvements? Non-existent. Reduced UAC harassment? Nope. In fact, aside from correcting a few well-publicized flaws, SP1 does little to improve the lot of Vista users. It is what it is, so stop waiting for a miracle and either suck-it-up or switch OS.
- Live Anything
Embrace and Extend lives on! Thanks to Office Live, we now have access to a half-baked attempt to marry the Web 2.0 and Microsoft's productivity suite. Unfortunately for the folks from Redmond, Live Documents does it better. I just love it when a little guy makes a fool out of the 800lb gorilla.
Happy New Year!
RCK
Posted by Randall Kennedy on December 29, 2007 08:30 AM
October 01, 2007 | Comments: (0)
Microsoft SoftGrid PM Hints at Future
Recently, I stumbled across a gem of a post by Chad Jones, Product Manager for SoftGrid at Microsoft. In it he touches on a number of topics related to the future of SoftGrid, what to expect from the next version and also how the technology will eventually be merged with other Windows Server components. Here's the link to the thread (hint: scroll down to see Chad's reply).
Some highlights:
1. SoftGrid 4.5 will enter beta in a few weeks (Q4 2007). From the sounds of it, this will be the last stand-alone version they release before absorbing the server components into Windows Server System, etc.
2. Version 4.5 will include a new, "lightweight" streaming implementation that doesn't' require Active Directory or SQL Server. This version will complement the existing SoftGrid server components as well as the new MSI packaging tool, providing 3 ways to deploy SoftGrid-packaged applications.
3.Microsoft will complete its integration of SoftGrid when it releases Microsoft Systems Center Configuration Manager (MSCCM) R2 sometime next year. Chad is promising "deep" integration with MSSCM, whatever that means.
Analysis:
I've been saying all along that Microsoft acquired Softricity in order to "strip mine" the core components. The MSI utility - which allows organizations to decouple SoftGrid-encoded applications from the streaming back-end - was the first salvo.
The revelation that they are developing a "lightweight" (i.e. pure-play, web-based) streaming server, one that allows them to deliver virtualized applications without all that AD/SQL baggage, simply confirms my prediction that they will eventually leverage SoftGrid to deliver a subscriptions-based Office product.
Finally, the hint about "deep" integration with MSCCM validates my assertion that Microsoft would ultimately dump the existing SoftGrid UI (which is klunky and rather ill-conceived) in favor of a fresh start under one of their existing management consoles.
Looks like I'm 3-for-3 in the SoftGrid predictions department!
Posted by Randall Kennedy on October 1, 2007 03:29 PM
September 24, 2007 | Comments: (0)
Microsoft's Virtualization Endgame
Big companies often acquire small companies and their products for the underlying technologies, and Microsoft is no exception. The strategy is particularly effective when the target is one of a handful of players in a nascent product category. This was the case last summer when Microsoft acquired Softricity. And though few of us knew it at the time, this acquisition – above all others – could leave the most indelible mark on the Windows product roadmap.
Note: For my detailed take on the current crop of application virtualization solutions, check out my new product roundup at the Test Center: On the road to the virtual desktop.
To fully understand Softricity's impact you need to roll back the clock to 1995. The Web was brand new, and a company named Netscape was giving Microsoft fits with its Navigator browser, and threatening to level the playing field through Web technologies. Microsoft, of course, fought back – first by matching Netscape's moves, then later by embracing the Web in its own, semi-proprietary way through ActiveX and Windows Internet Information Services.
We all know how that story turned out. However, few realize how profoundly this rush to embrace the Web altered Microsoft's DNA. Some of the more forward-looking company executives saw an unlooked for opportunity in the (soon to be) ubiquitous information superhighway. They imagined a world in which software was delivered much like a utility service, with subscribers instead of licensees. Already disenchanted with the upgrade treadmill and its unpredictable revenue streams, these executives envisioned doing away with shrink wrapped releases and the hassles of convincing customers to shell out their hard earned cash every 18 to 24 months. Utilities, they reasoned, don't have to worry about meeting revenue projections, and Microsoft wanted to join the club.
Of course, the folks from Redmond weren't the only ones to realize the potential advantages to what would become known as the "software as a service" model. Yet for all their enthusiasm, SaaS proponents were stymied by the limitations of technology. Early Web environments were simply too primitive to provide the kind of rich, interactive experience that users had grown to expect. This was especially true for the desktop where Microsoft had long established Office as the de facto standard.
Fast forward to today and the situation remains mostly unchanged. Microsoft is in the midst of another major upgrade cycle, and – just as in the past – Redmond is struggling to convince users to shell out for the company's latest and greatest. To their credit, Microsoft executives have managed to minimize their risk over the years by ensuring that virtually every new PC sold ships with the company's shiny new bits in place. However, this model still is subject to the whims of IT buyers, many of whom may balk at the idea of ripping and replacing their entire desktop stack just to use the new versions of Windows and Office.
To quote the title of one of Jack Nicholson's least inspiring film performances: "Something's Gotta Give." With broadband internet access nearly ubiquitous throughout the developed world, the foundation is in place for delivering a truly rich application experience via a hosted, subscription model. And with its acquisition of Softricity and the company's SoftGrid application streaming platform, Microsoft finally has the Internet-compatible distribution vehicle it's been looking for – not to deliver a watered-down, "Web version" of its applications, but the real deal. Think full-blown Microsoft Office (or Dynamics, or any number of Microsoft Games), with all the bells and whistles, streamed seamlessly to your PC from a series of massive Microsoft server farms.
With SoftGrid, Microsoft has the technology portfolio to enable its SaaS endgame strategy. The first salvo was fired at the VMworld conference when Microsoft announced it was decoupling the SoftGrid Virtual Application runtime client from the back-end streaming server (see my review of SoftGrid 4.2 for more on this development). By doing so, Microsoft is positioning the SoftGrid environment as a distribution mechanism on par with its Virtual Hard Disk (VHD) format. Soon you'll be able to "test drive" Office, et al, via this new, highly portable medium.
But test drives and demos are only the tip of the iceberg. Once Microsoft has successfully driven the SoftGrid format into the mainstream expect the company to slowly begin phasing out traditional delivery modes. That means no more Office CDs/DVDs with a thousand installed components – an MSI-wrapped SoftGrid OSD file is the future distribution model for virtually all of Microsoft's non-server applications. It's the perfect segue towards the "holy grail" of software development: Uninterrupted, subscriptions-based usage, i.e. utility computing.
Don't expect the move to a hosted, subscriptions-based model to signal the end of Microsoft's annoying anti-piracy technologies; the likes of the universally loathed Windows Genuine Advantage (WGA) will live on. As the folks over at Thinstall are happy to point out, SoftGrid is far from a secure distribution platform. In fact, once an application has been cached locally it's almost trivially easy to isolate the bits and extract a working copy. For Microsoft to protect its intellectual property, it will need to employ some sort of validation logic beyond a simple subscriber login. However, the implementation will likely be more granular than the current WGA model, possibly woven into multiple components within the virtual image. Suffice to say that Microsoft won't be releasing a hosted offering until it has worked out how to effectively secure it against piracy.
And what about third party software developers? Although Microsoft may make SoftGrid available as a Web-friendly distribution medium, it will no doubt keep details of its WGA-successor close to the vest. Thus developers will be faced with a choice: Deploy using a potentially insecure distribution platform, or seek out alternatives from the likes of Symantec and Thinstall. In fact, with Microsoft sure to sew up a sizable chunk of the market for enterprise deployment with SoftGrid, one of the more lucrative niches for competing technologies may be in the commercial distribution market. But no matter how you slice it, the future of Windows application deployment and on demand delivery belongs to application virtualization.
Posted by Randall Kennedy on September 24, 2007 08:20 AM
August 15, 2007 | Comments: (0)
Another Hit from the VMware Crack Pipe
It's been a couple of weeks since my last update. I've been transitioning to my summer home on Mauritius (always an arduous task, but more so with wife, 2 children and 4 notebook computers in tow), however, I couldn't pass up an opportunity to enjoy another hit from the VMware crack pipe. This time the fantasy comes courtesy of company co-founder, and chief "scientist," Mendel Rosenblum. Mr. Rosenblum used the podium at LinuxWorld to wax poetic about what he believes is the impending death of the operating system as we know it. In his version of the future, customers will move away from the traditional OS model and towards one dominated by "virtual appliances," where an individual application and its supporting OS are "merged" into a single, isolated component that runs atop a virtualization layer (i.e. VMware’s classic hypervisor scenario).
My response: Man, am I glad this is all just a hallucination!
But what if it weren't? What if every application came tied to its own, proprietary implementation of whatever OS its developer decided was ideal? What if we actually had to support and maintain such an albatross? How about in the desktop space?
Having to master Windows, Linux and (maybe) MacOS, with their myriad complexities and runtime quirks, is hard enough. Being forced to tweak a "merged" virtual machine image for each application is enough to make me re-consider that career waste management (at least there I'd know in which direction all the “"bits" are supposed to flow).
Frankly, I'm glad this is all just a smoke-induced delusion. The damage that such an archaic compute model would inflict on innovation would be fatal. The majority of advances in client computing hardware - the advanced graphics, rich multimedia and diversity of plug & play peripheral devices - can be traced to a single driving force: Microsoft Windows running natively on the bare desktop iron. None of these technologies works well with virtualization, which is why IT isn't yanking desktops en masse and replacing them with Mr. Rosenblum's pipe dream. They’d have a riot on their hands.
In the end, I think we all can appreciate VMware's dilemma: When your only tool is a hammer, every problem looks like a nail. However, I really wish they'd dial-down the hyperbole a bit. Windows, the bloated, insecure, monolithic OS that everyone seems to love to hate these days, isn't going anywhere. After all, Microsoft needs *somewhere* to deliver all of those SoftGrid-enabled subscription applications it has lurking up its competitive sleeve.
Posted by Randall Kennedy on August 15, 2007 02:38 PM
July 13, 2007 | Comments: (0)
Virtual PC 2007: The Great CPU Gobble!
Ever since VMRC Plus came out I've been spending more and more time with Virtual Server 2005 R2 SP1. As you may recall, I extolled the virtues of VMRC Plus a week or so ago, and since then I've been reaping the benefits of VS 2005 as a testing and development environment.
One of the unexpected side effects of moving from Virtual PC 2007 to VS: Lower CPU utilization. For some reason, when I load up my test scenario (two client VMs collecting data and uploading to a single server VM) in Virtual PC 2007, the virtualpc32.exe process hosting the scenario chews-up 50% or more of the available CPU cycles on my dual-core workstation. By contrast, when I load up the same scenario under Virtual Server, the VMs chew-up almost no CPU, which is what I would have expected given that the workloads running within them are very light (i.e. Task Manager inside the Guest OS sessions shows nearly zero CPU utilization).
I'm going to forward my findings to Ben Armstrong (i.e. the "Virtual PC Guy" at Microsoft) for analysis, however, I already have a theory on the source of the excessive CPU utilization. For starters, it seems isolated the client VMs, both of which are running a high-resolution monitoring agent (DMS Clarity Metrics Tracker). This agent makes frequent calls to PDH, WMI and the Registry, and I'm guessing that Virtual PC 2007 is generating a lot more overhead when processing these calls than Virtual Server. The agent also uses a set of high-resolution timer objects that likewise seems to give Virtual PC fits.
As a simple test, I tried disabling the agents on each VM. CPU utilization for virtualpc32.exe immediately to dropped to below 10%. Case closed.
Bottom Line: For testing applications that use high-resolution timers, or that make frequent calls to certain system libraries, Virtual Server 2005 R2 SP1 does a much better job of handling what should normally be a very fast, lightweight operation. And with VMRC Plus, you don't have to sacrifice usability in order to reap the rewards of better real-time application support under Virtual Server.
Posted by Randall Kennedy on July 13, 2007 09:50 AM
July 04, 2007 | Comments: (0)
Virtualized Vista Sucks on Linux, Too!
Last week I wrote about the poor performance of Windows Vista under the VMware Workstation and Microsoft Virtual PC environments. I noted that Vista was slower than it should be under virtualization, almost as if something - a poorly written kernel driver or bug in the scheduler - was dragging the OS down during benchmark testing.
Since my original tests were isolated to virtual machines running atop a Windows Host OS (Vista 64-bit), I decided to shake things up a bit and repeat the scenarios using VMware Workstation 6.0 running on Ubuntu 7.04 "Feisty Fawn." The results only served to deepen the mystery.
As with Vista under Windows-based VMware, the Vista-on-Ubuntu VM was slower than it should have been. But whereas the former configuration was as much as 50% slower than its equivalent native performance delta – as measured between Windows XP and Vista on bare hardware – indicated it should be, the Vista-on-Ubuntu scenario showed a 93% slower result. In other words, if the delta between XP and Vista running natively is ~2x, and the delta for the same OS configurations under Windows-hosted virtualization is ~2.5x, then the delta under Ubuntu is ~3x.
All things being equal, the deltas between these test scenarios – non-virtualized Windows XP vs. Vista on hardware, virtualized XP vs. virtualized Vista on Windows, and virtualized XP vs. virtualized Vista on Ubuntu – should be identical. However, in each case some mitigating factor is causing Vista to run more slowly, as measured by the completion times for the OfficeBench test script, than I would have expected it to give my experience benchmarking the OS natively. And this factor seems to have an even greater impact when running atop a Linux-derived Host OS.
Frankly, this is the opposite of what I expected when I set out to repeat the tests under "Feisty." I figured that the lower overall footprint of Linux – i.e. the whole “runs well on older systems with limited RAM” claim – would have contributed to a better showing across the board. In fact, the opposite now seems to be the case:
When it comes to the performance of virtualized Windows XP and Vista, Microsoft's newest OS mops the floor with Linux...or at least with the Ubuntu distribution of Debian. Go figure!
Posted by Randall Kennedy on July 4, 2007 08:51 AM
July 02, 2007 | Comments: (0)
Virtual Server: Now More Accessible Than Ever
I've always been a big fan of Microsoft Virtual Server on the desktop. A few weeks back I noted how VS was a powerful, free alternative to Virtual PC and even VMware, especially for users who need more robust CPU scheduling than the anemic, single-threaded VPC provides. However, VS has been plagued by a rather quirky, web-based management UI that forces you to run IIS in order to create and manage VMs. From a pure usability standpoint, Virtual PC always won hands down - until now.
Enter VMRC Plus. The Virtual Machine Remote Control Client Plus utility is a rogue internal applications that Microsoft developers cooked-up to circumvent the very UI issues I was lamenting above. Featuring a simple, if somewhat cluttered, user interface, VMRC Plus lets you do everything you could with the web UI but implemented as a native Windows application. That means real menus, real dialog boxes (for selecting local files, like ISO images), and all of the richness of a "fat client" GUI.
Thinking that their customer base might appreciate what had become a hot item internally, one of Microsoft's consultants decided to polish up the tool so that the company could make it available for external consumption. The result is now avilable as a free download from Microsoft's web site. Click here to grab your copy of VMRC Plus.
Some of the more useful VMRC Plus exclusives include:
1. Direct control of local or remote instances of the Virtual Server service. IIS and IE browser are no longer required!
2. Tabbed interface to quickly jump between Virtual Server hosts and guest VMRC sessions.
3. Reusable saved states: this feature allows users to preserve a particular saved state and return to that state at any time.
4. Multiple guest selection supported for startup, shutdown, save and display operations.
5. Browse button navigation for media, hard disk images, ISO images, .VMC files, etc.
6. Drag and Drop support for .VMC files, ISOs images, VHD and VFD files.
7. Resizable desktop support for guests running Virtual Machine Additions (maximize VMRC window supported).
8. Limited cut and paste of text from host to guest (only).
9. A built-in utility to take JPG screenshots of running guests. Useful when filing bugs.
10. Built-in error notification with Virtual Server eventlog viewer.
11. A Virtual Networks Manager and Virtual Disks Manager that cover all features.
12. Keyboard shortcuts (e.g. Ctrl-S to save state a guest).
13. Create multiple guests at once.
14. Create guest from parent (or multiple guests)!
15. Automatic reconnect to a designated Virtual Server host.
16. Toolbars in both Guest and Console Manager for quick access.
17. Unlimited number of guests.
18. Maximum of 32 Virtual Server hosts.
19. Sorting on columns of guests so you can sort based on status and multi-select.
20. Automatic detection of Virtual Machine Additions and notification.
21. Detection of Virtual Server 2005 R2 SP1.
In practice, the only thing I found missing from VMRC Plus is support for drag & drop between host/guest. And since this is a limitation of the VMRC client (which works more like an RDP session than a Virtual PC window), I can't really fault them for omitting something that was technically very difficult, if not impossible, to implement.
Bottom Line: VMRC Plus is a welcome new option for managing Microsoft Virtual Server environments. And while casual users will still find Virtual PC to be the "friendlier" way to VM under desktop Windows, hard core power users and developers will relish the chance to leverage Virtual Server's robust multithreading from within a more palatable UI.
Posted by Randall Kennedy on July 2, 2007 11:03 PM
June 22, 2007 | Comments: (0)
Vista + Virtualization = Poor Performance
Everyone knows by now that Vista is slower than Windows XP. In fact, my own testing shows it to be roughly twice as slow on the same hardware (see our upcoming Test Center study for more details). That's because Vista is a far more complex operating system, with many additional features and background services that simply don't exist under XP.
What many users don't know, however, is that this gap widens even further under virtualization. Fire-up Vista under VMware or Virtual PC and you'll find that the delta is more like 3x - i.e. tasks take over three times as long to complete under virtualized Vista as they do under virtualized XP.
Let that preceding statement sink in for a minute. In real world terms it means that applications are taking as much as a 50% greater performance hit from being virtualized than would be expected given the aforementioned 2x delta in a native, non-virtualized comparison. As a veteran IT professional, I would expect to take a hit moving to Vista - just not one that's so out of whack with the established norms.
Clearly, there's something going on here that makes Vista particularly difficult to virtualize. I have some theories, however, at this point I can only go on what I've observed. And that is:
1. When testing Vista on a Dell OptiPlex 745 with 4GB of RAM, the performance delta - as measured by the OfficeBench test script - is roughly 2x. The script took twice as long to complete under the Vista/Office 2007 combination as it did under Windows XP/Office 2003.
2. When repeating these tests under VMware Workstation 6.0 and Virtual PC 2007, running on two different hardware test beds (the aforementioned OptiPlex and an XPS M1710 laptop), the script took 3x longer under Vista.
3. All of the tests were conducted using OfficeBench, which is part of the Clarity Studio test framework that is freely available via the exo.performance.network web site (www.xpnet.com).
For the record, I was skeptical of the original results, so much so that I re-ran the VMware scenarios multiple times and then confirmed them against a different installation running under Virtual PC 2007. I also presented my findings to both VMware and Microsoft, but neither could explain the phenomena I observed.
Bottom Line: Vista is significantly slower under virtualization than it should be, and I'll be damned if I know why.
Anyone have an idea what might be going on here? Anyone? Bueller? B-u-e-l-l-e-r?
RCK
Posted by Randall Kennedy on June 22, 2007 07:26 AM
April 10, 2007 | Comments: (0)
Often when I review a product for the Test Center I'm asked to re-confirm my findings by one of the participating vendors. This is especially true when benchmarks are involved. For example, in the case of my recent roundup of desktop virtualization solutions (see the review), it was VMware that requested I re-test their product, VMware Workstation 6.0, using a later pre-release build (my original findings were based on the Beta 3 release).
Running the tests a second time Using build 42757 (a release candidate) indeed showed a change, but not the one VMware was likely hoping for. Client/Server database performance, as measured by the Clarity Studio ADO Stress workload object, slipped by nearly 3% (2.41 seconds/transaction vs. 2.34), while local messaging application performance, as measuring by MAPI Stress, took an 11% hit (5.22 seconds/transaction vs. 4.70).
Note: You can grab the entire Clarity Studio toolset for free at www.xpnet.com.
It's worth mentioning that my original scores from the Beta 3 build were achieved after implementing an unofficial "tweak" to disable the pre-release debugging code (basically renaming the bin-debug folder and copying the bin folder in its place). This, in turn, may have artificially inflated the Beta 3 results making VMware Workstation 6.0 look faster than it really was. I'm taking this opportunity to set the record straight and to finalize my benchmark testing of VMware Workstation 6.0.
Posted by Randall Kennedy on April 10, 2007 02:00 PM
April 03, 2007 | Comments: (0)
The Virtual Desktop crowd is salivating at Microsoft's latest tweaking of the Windows Vista licensing model. Dubbed Windows Vista Enterprise Centralized Desktops (VECD), this modification allows organizations to deploy Vista as a virtualized OS image on a centralized server. It also allows Software Assurance customers to deploy the new OS across diskless PCs and thin clients - a model that a niche segment of customers has already been doing with Windows XP (though it's nice to know that Microsoft is officially endorsing this approach).
Of course, it didn't take long for VMware to jump all over this new development. After all, it plays right into their Virtual Desktop Initiative (VDI) efforts - a.k.a. their master plan to "rule the IT world" from end-to-end. However, one comment by Microsoft's Scott Woodgate (in an interview for a competing publication that shall remain nameless) seemed to put a damper on the whole party.
He said, and I quote, "...with VECD, I would be spending up to 10 times as much on hardware to support the same number of users as I would on Terminal Services."
It's a sentiment I've heard echoed by some of my largest customers (various global financial services companies that shall also remain nameless). Sure, they could provision one of their data center boxes to run a VDI scenario - if they wanted to dedicate many thousands of dollars in server hardware to support just a handful of users.
All of which brings me back to one of my core themes in this blog: VDI, like terminal server, is a niche solution for certain non-mainstream compute scenarios. Just as we didn't abandon the PC in favor of server-based computing, so too will VDI fade as the combination of high hardware costs and the ongoing evolution of traditional "fat" client environments erodes VDI's already limited appeal.
Posted by Randall Kennedy on April 3, 2007 02:47 PM
March 30, 2007 | Comments: (0)
I received some feedback recently over my coverage of Parallels Workstation for Windows 2.2 (see the Test Center review). Specifically, I was criticized for describing Parallels as a "hypervisor-based" solution, a term echoed by the company's own marketing collateral. However, there seems to be some debate as to just what constitutes a "hypervisor" and whether or not the approach taken by Parallels truly qualifies.
According to Wikipedia, there are two types of x86 hypervisor:
1. A Type 1 hypervisor runs as the primary control layer and occupies Ring 0 of the Intel x86 CPU privilege model. In such a scenario, host operating systems - which would normally run in Ring 0 on bare iron - are pushed-up to Ring 1 (the next lower privilege level) when running within a Virtual Machine controlled by the hypervisor (which is now essentially the new OS kernel). Examples include XEN and VMware ESX Server.
2. A Type 2 hypervisor runs atop an existing host OS and manages the interaction between guest OS and the relevant host OS services and devices. In such a scenario, the hypervisor may or may not install kernel mode components - drivers, pass-through devices/services - however, the host OS remains in overall control of the underlying hardware. This is the model used by Virtual PC, VMWare Workstation/GSX Server, etc.
From what I can glean from the limited technical documentation available, Parallels employs a classic Type 2 hypervisor with a few additional kernel mode drivers to eliminate some of the slower aspects of VM-to-host OS integration (binary translation, etc.) And since the Parallels kernel-mode code is effectively in control of the entire system when executing (as is any Windows kernel mode device driver), the net result - managing the bare iron to accelerate VM performance - remains the same.
Note: Parallels never claimed to have implemented a Type 1 hypervisor. Rather, the confusion seems to surround the definition of the term itself and the fact that any mainstream virtualization solution can claim to be "hypervisor-based" per the above definitions. So while some may object to Parallels use of the term, they're really just splitting hairs.
Posted by Randall Kennedy on March 30, 2007 10:00 AM
March 24, 2007 | Comments: (0)
Thinstall Hints at Future Plans
For months now I've been arguing that the days of the traditional, locally-installed Windows application are numbered. Its replacement: A combination of ubiquitous connectivity and streaming, application-level virtualization - two irresistible forces that, I believe, will put the final nails in the fat client application coffin.
Microsoft gets the message, which is why they acquired SoftGrid: To enable them to deliver the next version of Office via secure, web-based subscription. It's also why Symantec gobbled-up Altiris. In fact, practically all of the major virtualization players have been overtly positioning themselves to be the enablers of the coming distributed, subscription-based compute model - all, that is, except one.
The folks at Thinstall have been notably quiet about their long-term plans. For example, I know on good authority that they intend to compete directly with Microsoft in the web-based application delivery arena. However, they have yet to publicly acknowledge their overall strategy. Still, if you do a little digging you come across hints at their intentions - for example, their recent decision to post a Thinstall virtualized copy of OpenOffice 2.0 on their web site.
I, for one, was happy to have the opportunity to play with OpenOffice outside of a Virtual PC or VMware session, and to be able to do so without the messiness of a local install. But I also can help wondering if this isn't the precursor to a far more ambitious effort - one that might even include a VMware-like library of virtualized "appliance-like" applications - in the future.
Regardless, it was nice experiencing OpenOffice locally thanks to Thinstall.
Posted by Randall Kennedy on March 24, 2007 09:00 AM
March 17, 2007 | Comments: (0)
Stealth Tech: PortableApps.com
I love stealth technologies. The chance to stumble across a product or service that has flown below the radar, yet has the potential to shake-up conventional wisdom, is one of the fringe benefits of penning a topical blog like this. So I was filled with glee at discovering PortableApps.com.
Portable Apps are mainstream open source or freeware Windows applications that have been repackaged to be portable. By portable, I mean the application's code and data storage requirements have been restructured to make it possible to execute them from a portable media source - a USB key or similar device - with little or no footprint on the hosting PC.
Why I find this exciting is because of the technolgy's potential to evolve beyond merely altering applications for portable device deployments. The truth is that the Portable Apps model could be applied to any scenario that requires a minimal footprint deployment with isolation of application code and data. In this context, Portable Apps becomes the equivalent of an open source Thinstall.
If you're currently considering an application virtualization solution, it behooves you to take a quick look PortableApps.com. Download the free Portable Apps Suite demo package. It may be all the application isolation you need.
Posted by Randall Kennedy on March 17, 2007 10:00 AM
March 12, 2007 | Comments: (0)
An Overlooked Alternative: Virtual Server 2005
With the launch of Virtual PC 2007 Microsoft made it clear they have no intention of competing in the desktop virtualization market. The product's lackluster performance and limited feature set will no doubt drive the remaining hangers-on firmly into the VMware camp. However, savvy users long ago discovered a viable alternative, one that doesn't require leaving the Microsoft fold.
I'm speaking of Microsoft Virtual Server. Though not marketed as a desktop virtualization solution, Virtual Server 2005 R2 (SP1) works quite well under Windows XP and Windows Vista. It provides all of the basic VM capabilities, including Intel and AMD hardware virtualization support, and is surprisingly easy to use/configure (provided you don't mind a web-based user interface).
It's also blazingly fast. As I'm writing this I have a Virtual Server VM installing Windows Vista in the background. The process is clocking in at less than 30 minutes, which is competitive with some high-end PCs. In fact, Virtual Server's performance is so impressive I decided to use it as the baseline reference platform for my upcoming desktop virtualization roundup.
Oh, and did I mention it's free? Definitely worth checking out. Grab the download and take it for a spin.