For now - check out http://globalfoundationservices for lots of great information we share openly about our data centers.

Also, I'll be speaking at one of our partners conferences in a few weeks:   The Experts Conference, or TEC, put on by Quest.  http://www.theexpertsconference.com/us/2011/   I'll be sharing a pretty thorough overview of our approach to data center design including construction of our modular datacenters...

Here's a ton of info on Microsoft Global Foundation Services (GFS)

Also wanted to share two links:  a Burton / Gartner arcticle (if you have rights) on making the transition from traditional datacenter to private cloud, and also a link to the GFS team blog

Good question.  And 90% of the time my answer is the same:  Start with Operation Manager. 

Here’s why:  you need to have an excellent understanding of what you have and how it’s behaving before you can be successful with anything else.   Use the building analogy:   You’ve got to have a rock-solid foundation before you even THINK about putting down a floor or standing up a wall.  Let’s take AD for example, something many people have and can relate to.  (and this does not indicate any problem with AD.  AD is rock solid; tens of thousands of customers love it.  But I’ve been working with AD for 10+ years and I still make configuration mistakes)

You want to start putting up a SCCM infrastructure to manage workstations.  If your AD isn’t 100% healthy, you’re going to have issues.  If you try a Windows 7 deployment and you have DNS lookup issues, you’re going to have problems.  Even if you start trying to manage those Windows 7 workstations with GPOs and you have an AD replication island you didn’t know about, you’ve going to have issues. You need to make sure you’ve got the solid foundation first. Can you tell right now if your AD is 100% solid with nothing going wrong?  Has it been performing well for past 3 monhts?  Having any external DNS lookups?   Are you proactively finding and resolving issues before anyone calls helps desk?  

I’ve had customers tell me SCOM have saved them hours in both troubleshooting and addressing potential minor issues before they become service-impacting issues. 

Now, don’t think this you have to stop whatever project you’re doing and do SCOM first.  This does not have to be a full-blown SCOM project with 4 months of planning:  Start with it small. Get a SCOM instance up and import the WS08 Base and AD management packs, let it marinade for a while with the out-of-the-box settings, and it’ll start letting you know if something isn’t right.  Take advantage of the knowledge from the guys that wrote AD, the guys that live it every day in Premier services.

You may already have a monitoring solution already, and that’s good – something is better than nothing.  But SCOM gives you so much more than any other solution, especially as you start mixing physical and virtual serers.   And if you go with other System Center products down the road, remember they all take advantage of each other (SCOM + SCVMM and PRO Tips for example for moving VMs around from host to host, Service Manager as your ITIL/MOF CMDB, etc.)

Well, a few months ago I had the opportunity to work with one of our ISVs on creation of a management pack for their application.  This ISV wanted to have a MP available as an additional value-added selling point for going with their app.  The goal for them was to expose out the most important parts of their application behavior and performance, so IT department could increase their manageability of the app -- and ultimately provide a better experience for their customers.  I thought this was amazing; that an application developer wanted to create and ship a MP with their app and use it as competitive advantage. 

This engagement was very interesting -- it was a bunch of developers from their side talking to me, an infra guy.  (Almost reminded me of travelling to a non-English speaking country.)  Luckily I had a teammate who knows application development and  coding best practices to be my interpreter.

For those that aren’t totally familiar with what a MP is, it’s all the knowledge about what an applications health looks like.   It basically watches for those “outputs” thrown from the OS, the app, the overall environment…  and the MP is the knowledge about what to do when it sees things.   This is where you put all the events to watch for, and what you’d tell an operator to do or investigate if they were to see a certain operating condition (like an error in the event log, or a service stopping for some reason)

As I was getting ready for their visit and brushing up on my developer talk, I had a tough time locating any process or flowchart written for an application developer wanting to create a management pack.   So, I thought I’d share the approach we took during this engagement, and share our experiences in order to help any other application developer get started.  If you have any ideas for improvement, I’d love to hear it.  (and I assume you have a basic understanding of what SCOM is, what a monitor is, etc.)

1-      We started by identifying what exactly needs to be monitored.  Begin with thinking about a “oh no, this is bad”, “this shouldn’t be happening”  or “um, this is kind of bad” approach.  Map these out; don’t worry about getting everything, just get the biggies out there.  You can add more later.   We call these Health Monitoring events.

2-      Then we figured out what needed to be watched for overall performance and trending.  Now, in order to answer questions like “what has been my uptime for this app over the past 3 months”  you’ll need to expose out info on a consistent pattern – like every 15min. We call this performance monitoring.

3-      Once we figured out those two models, we started by understanding how we could get answers to these.  In SCOM speak, these are called Monitors.  There are lots of monitors, for Event logs, for WMI events, and so on.  Bottom line to SCOM:  if its somehow exposed out, there’s a monitor for that.  So, basically you need to determine if these two criteria (health monitoring and performance monitoring) can be met by existing out-of-the-box monitors available in management packs.  For example: SQL MP, Windows Core services MP, etc.   For this dev team, it turns out that about 80% of what they wanted to monitor already existed in those MPs.  Cool.

4-      We started by authoring a new MP in the SCOM console – the easiest way to create rules against those monitors.  This got them almost all the way with what they wanted.

5-      Then when they were ready, we went for the SC Authoring Console – the more advanced way to author management packs.

As it turned out, many of the key things they were interested in monitoring for their application were addressed by standard management packs in SQL, Dynamics, and Windows Server.  They added and tweaked a few of the core value with overrides, and added some of their own services monitors.

At the end of four days, they were well ahead of where they thought they’d be in management pack creation.  But even more important – they learned a very important lesson:  they weren’t exposing enough of their health characteristics of the application, thereby limiting how much could actually be monitored.  They went back with a MP, but better yet they went back with the knowledge about changes they needed to implement in code to create more robust abilities to event- and performance-monitor.  Once they mature in their MP development, I have a good feeling they’ll be able to do some really advanced health monitoring for this app.

Good luck guys!

If you use HP ProCurve Manager to manage your ProCurve switches, you know its a pretty powerful product.

There's a security feature that you need to modify if you want to have another workstation view the information from your ProCurve Manager server.  By default, no other workstations can connect to your PCM server, such as an Operator or extra monitor in your NOC.  If you install PCM on a workstation, and try to connect to the PCM server, you will receive a message:  ":  the server would not authenticate this client, you may need to upate the list of authenticated clients on the management server"

A search online and in the help yields nothing.  So, here's what to do...

On the server, locate the file ACCESS.TXT.  Edit this file.  It will probably be empty.

add a line of the address of the workstation that wants to connect, or even easier, add the subnet of workstations that you trust to connect (like 10.*.*.*)

restart PCM client on the workstation, and you should now be able to connect...  hope that helps.

 We deployed an interesting solution that makes me want to revisit the whole "stream the OS to the desktop" scenario.  With Citrix added on to Windows Server 2008, you can boot from an ISO image.  We use that here to avoid some conflicts with an existing demo, where we boot the ISO and start up a VHD, and then connect the user to the VHD for their VDI experience. 

But this got me thinking... if i can boot to an ISO, i'm in effect streaming the OS down to the workstation.  if i create my own special ISO image, and can send it down as the workstation boots, we can basically operate a diskless workstation.

So, now that the environment is up & running here, i need to spend some time with this to see how this could enable some solutions in highly unmanaged environments (retail scenarios, fast food chain restaurants, etc.).  Stay tuned.

(also known as "how i spent my christmas vacation")

 http://edge.technet.com/Media/Inside-the-new-Chicago-Microsoft-Technology-Center-MTC/

I'll try to get my deck posted somewhere.  If you're at the conference stop by and say hello...

RemoteApp allows you to publish out an application, as if it was installed locally on the users workstation but actually runs in the memory space of the server.  This is ideal for users that need access to an application every once in a while, or need it remotely, but doesnt require you to install anything on the users desktop.  And if the program uses a file extension, like .MPP for Microsoft Project files the client can actually recognize this and start the remote program on the clients behalf.

Now this is great if you are on the corporate network or VPN'd in.  What if you want this access for a person who only has https:// access?   Dovetail RemoteApp with Terminal Server Gateway, and you can enable this across untrusted networks. 

We ran this for a very large customer who wanted to proof-of-concept our Microsoft Dynamics applications, and it worked like a champ.  Now, we had to create a cert through Microsoft trust chain, which was the worst part about the process, but once setup it was hands off.  and *very* enabling!

I can really see this changing way applications are deployed and managed...

I’ve been wanting to look into VLAN identification feature of Hyper-V, and see how this can really benefit us and our customers.  After some reading and experimentation got it figured out and wanted to share my experience...

In the old days, I’d have a physical NIC for each VLAN and configure the switch port the NIC went into to be “hardcoded”  (untagged in HP ProCurve lingo) to a VLAN.

On the Server, I have it configured according to the common preferred practices:  one NIC on Management network used only for management, the others are are cabled to a port on our ProCurve switch .   Important note:  the physical NICs themselves are not VLAN tagged (i.e. properties of the NIC card), but the port on the switch that the NIC goes into is marked as TAGGED.