The PDC is over and I am home. I am taking a few minutes to be a little reflective and put down some thoughts.

1. The Acer PC was great. I heard people complaining about it not having more software (Visual Studio 2010 being the most common) or RAM but I have to say I was looking for a second machine for demos and this is better than what I was looking for. It makes up for no breakfast or attendee party. Of course now Microsoft has set an expectation and it will be interesting to see if they give out other goodies like a mobile phone at MIX (hint, hint…).

2. The roadmaps were shorter. In previous PDCs the roadmap slides seemed to go out to 3 or 5 years. Almost all of the roadmaps this year only went out 18 months. This felt more like a TechEd with the shorter timelines. Of course with how little in the 5 year time frame actually got delivered it is possible that Microsoft just realized that the information wasn’t as useful as they would hope.

3. Sessions were generally good quality. I went to more sessions than I blogged and for the most part the speakers were good and interesting. In the past it seemed they just grabbed some random employee to get up there and talk but maybe they screened the speakers or gave them some training.

4. Networking was even better than the sessions. I think the real reason to go to any conference is the learning. Some occurs from the sessions but more occurs from the conversations at lunch or in the hall ways. I have come to appreciate the need for this more as I work from home and don’t often get a chance to discuss things with my peers as often as I would like.

5. Focus on some general themes. In the past PDC seemed to be about anything that any product group wanted to announce. I didn’t see the breadth of topics represented this time. I think that was a feature of the shorter timeframe but it was nice to have focus on themes like SharePoint, Azure, Windows, and Visual Studio. It meant there were a lot of relevant sessions. The down side was that there were many times when I wanted to see 2 or 3 sessions in a given time slot.

I am looking forward to the next PDC and to see how things like Azure, “Dallas”, and Silverlight 4 evolve.

Local access infeasible so you can’t get access to any of the event logs or other tracing that you would normally do on a single server

SDK supports distributed monitoring & data collection for cloud apps.
Support Standard Diagnostics APIs
Built on top of Windows Azure Storage
The same infrastructure is used by Microsoft for their monitoring so they know it scales.
Developers are in control of what gets collected and when to collect it.

MonAgentHost.exe is the diagnostic monitoring piece that is doing the monitoring and is started by the fabric UI on the developer fabric

Used Cloud Storage Studio from Cerebrata Software to show off storage information.

WADDirectoriesTable and WADLogsTable store log information. The data in the directories data is the standard IIS log files that are in put into development blob storage.

The diagnostic monitor is a separate process and can do things like crash dumps and Windows Data Sources like event logs. Data goes into the local storage directory and is then uploaded to Azure storage.

The oldest data will age out.

System.Diagnostics.Trace to write information out to the logs. Reference Microsoft.WindowsAzure.Diagnostics and import the namespace.

DiagnosticMonitorConfiguration lets you set up the configuration information. Everything is buffered locally by default and not uploaded to the storage. You have to provide the storage configuration when you call the Start() method.

TraceListener added iinto the web.config by Visual Studio to allow the monitor to listen to the events and store them.

Microsoft.WindowsAzure.Diagnostics.Manager namespace used to write an application that will let him do on-demand uploads of data to storage.

If you are trying to catch crash dumps in a web role ASP.NET will most likely capture the data so you won’t be able to examine the dump. If you fail in the startup or in a worker role you will be able to get the crash dump information.

Turning on IIS tracing incurs a performance overhead and can not be turned on or off dynamically so you will need to upgrade your application to change the setting.

There is no automatic deletion of logging data from the Azure storage. You need to clean it up so you don’t pay for log entries you don’t need.

Data partitioned by the high order bits of the tick count so you can query on just that partition.

WMI is not supported natively but your role can reference WMI and log the information into an “arbitrary log”.

The role runs in the “Performance Log Users Group”. Soon IIS logs will be generated in the role’s local data directory.

Azure contains 5 main pieces:

1. Windows Azure Application Platform
2. SQL Azure
3. Windows Azure platform AppFabric
4. Azure “Dallas”
5. Pinpoint marketplace? (I came in late and only saw the screen for a second so I am guessing this was the 5th element)

Different roles on architecture slide:
Web Role
Dynamic Worker
Distributed Cache Worker
Partitioned Worker

Fundamentals:
Security
Performance and reliability
High availability
Scale out
Multi-tenancy

Service healing is available because the data is copied to multiple servers and if one goes down that can be detected, that instance shut down, a new instance spun up, and a message sent to the load balancer to start sending load to the new server.

New will be drives so you can map to Azure storage and use standard APIs to manage the data.

Coming soon:
1. Programming model – administrator privileges in the VM
2. Storage – user-selectable geo-locations for replicas
3. Service management – remote terminal server access to VMs

Note: These are my raw notes from some of the sessions I attend. Items in italics are my comments. Others are notes from the slides (or at least what I heard).