Keynote - Detail

As Colin has said I took lots of Notes. I also took some pics of the empty stage, to give an idea of the size

These were from the 1/3 nearest the stage - I reckon about 6k attendees could be accommodated.


You could really see the impact the flight ban has had on attendance as it was maybe 3/5s full at most. Hopefully there will be more attendees tomorrow now flight restrictions have eased.


Anyhoo back to the details...


Brad Anderson started off as the warmup for his boss Bob Muglia, Brad took us through some stats -
3 out of 4 attendees use SCCM, 80% of which are already on R2
2 out of 3 use SCOM, again 80% at R2
50% of SCOM users, take advantage of its hetergenous features to manage Unix/Linux
50% of attendees use System Center Virtual Machine Manager (SCVMM)
25% use App-V
10% are beta'ing System Center Service Manager


7 years ago MS first announced the Dynamic Systems Initiative, the first step on the path to Dynamic IT now they are making it a reality, the vision will continue to evolve.


He talked about things like the Lab management tool in Visual Studio 2010 which allows you to deploy your own test lab using Hyper-V and SCVMM.

Brad talked about Opalis, a recent acquisition, which has a Orchestration feature, which automates moving (virtual) Dev environments into production, with the whole environment available at once - no more multiple changes/over time.(Opalis is something we may own but is outside of scope of Connex - something for EUT to investigate further I think).


Next up was a demo and a pretty impressive one. There is a feature within Hyper-V which allows you to do a long distance live migration. This would allow failover between say ISB and MTO, with no user impact as the servers would migrate in a live state even over the huge distance.


Obviously MS are keen to push new parts of the System Center suite, they talked about the human workflow of change and how it can slow the process, System Center Service Manager (SCSM) can now do a change automation based on ITIL. System Center Data Protection Manager (SCDPM) has better functionality for backing up (Hyper-V) based virtual machines, down to individual files on VMs, not just a snapshot. Multi site clustering with Hyper-V and System Center products....and so on.


Actually they are making some significant improvements, I could well see it being time for Mars to assess Hyper-V certainly for Dev/QA environments as it is much cheaper than VMware and seems to be catching up in functionality and adding features VMware does not have.


Further areas of improvement will be including more Compliance management in SCSM and SCDPM. All about proving how they have and continue to deliver on their vision of Dynamic IT.


So Microsoft asks, "What next?"


The Cloud.


All the attributes MS have defined as Dynamic IT apply to the cloud. The Cloud they defined as just in time provisioning and scaling of services on shared hardware.


Why Cloud? Accelerates the speed and lowers the cost of IT. Brief definitions of Public/Private clouds (hosted/in house) and Shared/Dedicated (Shared with other customers/Service dedicated to you).


Microsoft is working to provide dedicated clouds with Azure in the future (Shared only now).

They are looking to deliver one platform, one application model and one management solution across all of - customer premise, partner cloud, MS clouds.


There are a few key enablers -

Hardware Model - Windows server is now 75% of all servers globally. MS now buy servers in 2000 server containers, they just plug in power, network and water. This is 10x more efficient than the process of provisioning individual/racks of servers. They are working with hardware partners on the learnings and expect to see smaller containers offered to end users in the future.


Application Model - This is a set of services delivered as part of the cloud - this reduces dev time, has increased scalability, higher high availability and greater flexibility. Again a 10x improvement over current methodologies to be faster to market. We need to understand that servers will fail, however, applications should not, the service should continue. MS are developing a new model language currently code named 'M' this allows a developer to build apps based on a model rather than traditional methods.


Operating Model - They have learnt a lot from running Bing! as a service with a small number of admins. They have taken this knowledge and built it into Azure and System Center to improve their products. They can now have 1 admin managing 1000's of servers! They suggest that IT jobs in this sector will evolve to provide a higher service, faster delivery etc. The underlying operating model enables this. They have seen (you guessed it), a 10x reduction in the cost of operations.


New features coming -
SCVMM v.Next will have the ability to manage OS/Apps that run across multiple machines (1 OS, multiple VMs - this I have not explained well, I'll try to find more info over the rest of the week). Applications are referred to as 'fabric layers'.#


Service Designer feature - allows you to deploy new services based on your templates (Customer logs call for more Oracle capacity - admin clicks on deploy Oracle service and the capacity is provisioned) basically you can draw the picture of your service in Visio and then SCVMM will deploy it....SCVMM will also scale up/down as the load increases/decreases as per your requirements. Great for end users, a nightmare for licensing compliance!


Server App-V use multiple apps on the one OS independently, SCVMM manages the underlying application fabric.


Greater control of offline patching - remember the app service must stay up, gives greater control and automation.


SQL Azure - running SQL as a service across 6 datacenters and 1000's of servers, provision of a new DB is as simple as clicking on a web page.


Finally, integrated monitoring between on premise and cloud - a SCOM management pack shipping later this year for Azure. The demo showed a diagram of the environment with hw onsite and cloud based, a simulated problem in the cloud alerted via SCOM allowed the admin to run a task to provision more capacity in the cloud. Again very impressive, but how do you manage the cost of this up/down scaling and the capacity required on standby? I think contracts will be very interesting!


My takeaways - we probably need to look at all the features of the products we have bought as part of Connex, not just focus on the immediate need (I think a common mistake in Mars and industry wide). There is much more to many of the tools that could allow for greater automation and much slicker operations just with a bit more upfront effort.


Secondly, we need to think more holistically and not just in our GIST silo's, products EUT are using will be more than useful to other teams, we need to ensure we highlight these to our colleagues (as we have done with SCCM and SCOM to Processing). This is probably a great example of where an Enterprise Architect function would be particularly useful - I think Chris Lane is going to be busy ;-)


I'll leave it here as I have another 4 sessions to blog, but you can find more info below.


For those that would like to watch today's keynote, it is now available here


Finally, tomorrow's keynote will be streamed live here from 8.30am Pacific