Infrastructure Optimization

This has been reported widely across the blogosphere, but if you haven’t yet linked this page, and you use Microsoft Deployment, stop what you are doing, and go here now.

The MS Deployment dev team has posted a Technical FAQ based on questions they have received via Support calls and email. Lots of amazing tidbits in here.

The team has pledged to update this site as new questions come in. Very bookmarkable!

Best Practice Alert – did you folks in the app-virtusphere know that Office 2k7 was not initially designed for virtualization?

I’m a bit bemused by the SoftGrid Blog team using that phrase “not designed.” I just wonder, with my tin-hat on, waiting for the black helicopters – did MS do this with intention? Will Biff go into rehab?

From the SG Blog – (link) Microsoft Office Support of SoftGrid Outlined

“Keep in mind that Office 2007 was not designed for virtualization, so there are some features that will not work in a virtualized environment. Find out more about best practices and limitations at www.windowsvista.com/optimizeddesktop.”

With intention? I don’t think so - along the sequencing path there have been a few add-ons to achieve packaging nirvana as well as updates to the AppVirt client. For instance to the AppVirt sequencer base image - we’ve added Windows Installer, .Net and VC++ redistributables as needed. Hey, even the AppVirt sequencer needs a nip and a tuck every once in a while - don’t hate it becuase its beautiful!

Reading on -

“Depending on the usage scenario, we recommend that customers install Office 2007 on their client machines to ensure access to all of the product features, and use virtualization for older versions of Office in cases where they need to run multiple versions of office at the same time. “

Best Practice-

This last quote is the “one ring” to rule them all that falls into “best practice.” We have found in our AppVirt deployments that its beneficial for long term support – from within your own enterprise and from Microsoft to follow best practice and install Office locally and sequence other versions for AppVirt delivery. We learned this lesson back in the version 2.x days on a two thousand plus seat deployment. While Office sequenced and streamed it was trial and error to find all the touch-points where other applications needed Office connectivity. It became very clear we needed to install locally and sequence and stream older versions.

Biff did go into rehab - my helicoptor awaits.

A follow-up to my article on Microsoft Deployment: Michael Niehaus, BDD developer and all-around cool guy wrote a “from-the-horses-mouth” history of Microsoft Deployment, for those of you interested in its evolution:

A Concise History of Microsoft Deployment

Microsoft recently released an updated version of the SysInternals AutoRun tool, which is absolutely in my daily toolbox when troubleshooting desktops. What runs at startup on your machine? Like detail?

It’s compatible with Vista now, and comes with a command-line version, for you automation folks.

Get it.

If you’re familiar with BDD, and especially BDD 2007, you know that Microsoft has been thrashing around a bit on names for the next version of their excellent Business Desktop Deployment Solution Accelerator. Well, they’ve finally settled on a very Microsoft-y name: Microsoft Deployment 1.0, available for download on Microsoft Downloads. I must say - this name shocked me when Michael Niehous revealed it. I would have placed good money down on a System Center appellation.

But, irregardless of the name, MD 1.0 (which doesn’t really have the same ring as BDD, but I swear - I’m done commenting on the name!) has improved upon the already excellent feature set of BDD 2007. Let’s have a look at BDD/MD’s evolution real quick though:

Microsoft Deployment History

Pictures are great and all but what does it all mean? Well, the bottom line is that BDD 2007 was perhaps the best example of a deployment tool from Microsoft that mortals such as me could use and understand. It’s also been a highly-successful tool for our customers at Getronics. So, as you can imagine, I have been following changes to this tool (critical to my livelihood!) closely, looking for anything that will fundamentally alter what we’re already doing with clients, or for opportunities to suggest new and better ways of doing things. Let’s look at a few that I’ve found so far. If you’ve found others - shoot me a comment!

Support for Deploying Windows Server 2008 (kinda!)

One welcome addition for testing labs (such as ours!) is the psuedo-support for easy deployment of pre-release versions Windows Server 2008. I say “psuedo” because MD 1.0 is not intended to roll out production 2008 machines, and well, to be fair, neither are the current versions of Server 2008.

Why I think this is great is that my fellow engineers and me are trying out different configurations on virtual hardware, and being able to re-image via PXE is super-sweet. And hey, the fact that it’s restricted to Lite-Touch (LTI) is fine with me. Who deploys servers via Zero Touch anyway? Google, perhaps.

Documentation Wheel Changes

One really obvious change from previous versions of BDD is how the Documentation Wheel works now. It looks the same, but there are some subtle changes here, bringing its terminology into the present and in line with current thinking on deployment.

MS Deployment 1.0 Documentation Wheel

For example, the rather confusing “Computer Imaging System” has been renamed “Image Engineering” which makes a lot more sense to me, at least. The documents behind all of these areas have been updated as well, and are well-worth reading. Documentation in BDD has always been kind of a two-edged sword: there’s a lot of it, but it seems to be organized in a difficult-to-fathom kind of way. That still plagues MD 1.0, but in all fairness, there is a lot to cover, and I’m not sure I could organize it any better. Suffice it to say, if you are just getting started with BDD/MD, set aside a day or two to read every “feature team guide” that’s available from the documentation wheel. The Gestalt of knowledge you acquire will make you better at implementing BDD/MD than any one of the documents alone.

Compatibility with SCCM 2007

A big one for us is compatibility with System Center Configuration Manager (SCCM), and the new ability to deploy XP, Vista and Server 2003 with MD and SCCM 2007. In addition, MD is completely integrated into the SCCM console, and shares the advanced task sequence features with its elder product. Task sequences can be built in, and imported from SCCM, which is really, really nice. Or, if you’re newer to the game, you can create “quick start” task sequences and packages from MD that are easily imported into SCCM.

MD 1.0 also extends the available actions in SCCM 2007 task sequences. Those of you who live inside the task sequencer, and have felt the raw power of these easy-to-explain, but hard-to-write-on-your-own automation tools will appreciate this.

This all means that MD and SCCM are ever-more tightly integrated, and if I had to guess, I’d guess that despite MS’s protestations to the contrary, MD will eventually be enfolded into SCCM, and will disappear as a standalone product. If anything, I would predict that if a standalone version does persist, MD will become “System Center Deployment Manager.”

Oh, and get this, BDD old-timers: you can now deploy to computers that SCCM 2007 does not see! This is kind of revolutionary, if you’ve struggled with getting all deployment targets into SMS/SCCM.

Multicast

Hang on to your hats, folks. But the biggest feature, IMO, of MD 1.0 is multicast. What’s that, you ask? If you’ve done large-scale deployments over heterogeneous networks, then multicast is just what you’ve been looking for. Some of the requirements are quite stiff, but if your hardware can handle it, its definitely worth looking into if you’re going to be doing significant hardware refreshes or upgrades.

Multicast is a technique in which one image is sent out from the imaging server, and is used multiply by the intended targets. In this way, multiple copies of the image are not being transmitted and overburdening your network.

There’s a new type of WDS (Windows Deployment Services) server called a Transport Server, and this is the new piece of the MD/WDS/SCCM universe that allows for multicasting. Read more on WDS in Server 2008 here. If you want to do multicasting, you’ll also need routers that are capable of it, and you’ll need to watch out for a bug: multicast only works with the boot.wim on Server 2008 install media. It does not work with the WIM from Vista install media. There are other caveats, but they are outside the scope of this article.

Small, but big changes

I don’t have the room to describe the full range of changes, but a few highlights are appropriate, before I go::

• Lite-Touch Installation: easy migration from LTI to SCCM 2007
• Support for multiple task sequence templates
• Task sequence rules can now invoke web services - great for keeping a running inventory of machines imaged, or for hooking into existing systems.
• Support for offline patching
• Support for language pack installation, both on- and off-line
• Windows Update integration
• Enhanced User State migration (stay tuned for a future article!)

In addition, if you aren’t ready to give up BDD 2007, the two products can now be installed side-by-side. I don’t see much reason to stick with BDD 2007 if you haven’t fully committed to it yet. If you’ve got a rock-solid BDD 2007 system, you may want to instance a lab before committing.

Overall, I’m liking what I’m seeing, and am looking forward to many, many deployments over the coming year.

While I’ve not finished parsing it, Tarry Singh’s article caught my eye. I don’t normally read all of the links that come in through the ITToolbox newsletter, but anything with “virtualization” in the title tends to give me pause.

Tarry’s article provides a (rather dense and obviously knowledgeable) collection of predictions for the future of virtualization in 2008. If nothing else, there are a bunch of great links to check out here, if you’re in the field.

Virtualization Predictions for 2008

Tarry - we’re watching too. Perhaps you’ll inspire us to make predictions of our own!

During a client-call yesterday, once again, we heard the “very many challenges” that data centers have with all of the issues that come with the incredible and never-ending demand for application services within enterprises.

A solution that’s well-tested and well-vetted for these ever-increasing demands for applications is the propping of virtual infrastrucutres. Nothing new here: just prop a virtual server or two, install the application, and voila - the enterprise is on its way, with no new hardware purchase.

But, taking this concept further - given this demand, and the repeatability of the process, how does this mode get streamlined? For a while we have known of a company that is offering the ability to create complete infrastructures in virtual design environment.

With their service, you draw in a Visio -like web-based program what infrastructure you want, how would you like to connect it, and such and then click on “build.”

In just a few minutes, all of your servers are created dynamically in the background. If you need more servers, you simply add to the visio-like diagram and they are added.

The company name is 3tera and can be found at: www.3tera.com

The demo of their control panel AppLogic can be found at: http://ns1.netclime.net/online-demo.html (also on their home page.)

The demo is particularly interesting (even for those of us that are not technical!)

And, in the interest of full-disclosure and above-board etceteras, the IO Team is not affiliated / associated with this company, but we know a good idea when we see one, and we believe this is the future of IT infrastructures.

With Google, Yahoo, Microsoft (and now Amazon.com) building monster (500,000+ servers) data centers, and touting grid-based computing to offer services, like those discussed above, you can see we are not too far from being there. Now imagine an organization built on a virtual infrastructure, with application virtualization all the way to the desktop. Now that’s powerful.

How does this affect the landscape of our future?

As more and more IT services get commoditized and virtualized, fewer technical folks will be needed on-site, and the need for good processes and people trained in technology will grow, resulting in business alignment along with managing demand. In our opinion, this the keystone for future IT departments.

Doing some research this week, I ran into a website called www.jumpbox.com which is a company that provides server-based software installations using Virtualization technology… Essentially they do all of the hard configuration and settings work, and provide a Virtual snapshot of the application ready for it to be used… They have several open-source server-based applications that are ready to be ran for free - Worth checking-out out for sure! –

As pointed out by my colleague Danny, and trumpeted in The Register, VMWare has recently purchased Thinstall.

http://www.theregister.co.uk/2008/01/15/vmware_buys_thinstall/

This puts some heavy muscle behind one of SoftGrid/AppVirt’s competitors. One can only speculate what the San Francisco to Palo Alto move will mean for the product, but personally, I hope they invest heavily in bringing Thinstall’s feature set into parity or dramatic competition with AppVirt. My colleagues here in Seattle and I are passionate SoftGrid-ites from back in the dizay and have, on more than one occasion, evaluated Thinstall as a potential technology for our customers. We found it somewhat lacking, but definitely “one to watch.”

It’s no news to anyone that VMWare and Microsoft are on a collision course for domination of the burgeoning virtualization market. If one believes the free market theories, competition will only make both products stronger, meaning that our collective hallucinations and passionate ramblings about “Software as a Service” and the Zero-Touch Desktop will only solidify faster. This move is a big win for IT departments considering application virtualization as part of their overall strategy for decreased TCO on desktops, more centralized administration, and self-service application provisioning. Oh, and if you’re among those IT departments not considering this technology, I’d love to know why not! I’m not being flip here: after working with this technology and envisioning the new avenues these products and methodologies open up, I strain to consider how this could be a bad thing.

The IOTeam will be watching Thinstall with a greedy hunger. We’ll let you know what we find.

Greetings, everyone! We’re the Infrastructure Optimization team at Getronics, and we’ve decided to start a blog. After years of internal discovery, we thought it might be a good idea to give back to the community, by sharing some of our thoughts and findings around the topic of Infrastructure Optimization.

Not quite sure what IO is? Well, we hope to change that. It’s an exciting field, with lots of facets, and if you’re in IT, you’ll be hearing more about IO over the coming months and years, if you haven’t already.

All feedback on our posts is welcome - we’re learning as well as teaching, and want to grow with the community.

We’ll be introducing ourselves shortly, but for now, we’re just happy to have our blog configured!

Cheers,

The IO Team