Notes from the SAF

1. Internet Service Bus
The speaker discussed the growing need for the internet to act as a conduit between multiple client footprints to the same data and services. No longer need to focus on internal ESB but instead treat the internet as THE service bus. Exposing interfaces that can be consumed by rich diverse clients is essential to service adoption.

2. Executive Conversation
Gates talked about, variation in ecological impact of scale up v scale out applications.
Social computing not taking off in professional space outside of the walled garden of the enterprise, adoption of SharePoint wiki and blogs highly sporadic
Computing power being distributed throughout the stack into nontraditional spaces, NICs, video cards etc, standards need to evolve to stop the development costs being so high with vendor tie in.
Role of enterprise architecture discussed as an enabler of business integration, an essential component of large corporations, failed as governance vehicles for small LBU.
Architecture frameworks, unable to adopt one across MSFT due to diversity of technologies, they all failed when attempted to be applied holistically.

3. Data driven approach to distributed concurrent software in the enterprise (REST)
Use of Microsoft robotics run time studio in high performance applications. Data driven premise is based upon the idea of a service is a living document, describing state. Services manipulate document and change the state. An application is a composition of services.
Robotics maps to high performance computing because of coordination of asynchronous components, spanning multiple compute nodes. Robustness and isolation key benefits of this approach. Two major components: CCR - coordination and concurrency library. DSS - distribution services framework. CCR can be used with DSS. MySpace is running on this core technology.
There is s Visual programming language/environment, pure dataflow diagrams used to bind the services together Runs on all .NET frameworks, mobile etc. Integrated Lightweight service hosting environment.
Document drives behavior; make it visible, this exposure makes orchestration simple. infrastructure deals with exclusive or shared access to state in document.
http://microsoft.com/robotics

4.Tafitit.com
Excelent discussion will blog this seperately.

5. Composite application architecture
Proposed a move to four tiers, add "productivity tier" manage the rhythm of the business here, because not all business rules are shareable. Hence one layer of shared business logic (application) and another that is specific to the workflow solution (productivity) lots of discussion about BizTalk being a good candidate in this layer.
Presentation, productivity, application, Data. Example of tiered technology supporting stacks office, SharePoint, soa/BizTalk, ERP
Why SharePoint? They had an interesting perspective that it is really a storage application that can manage lists, views, libraries, versions and deals with security. It provides an abstract layer on top of the data.

6. Amazon ECC, S3, SQS
SLAs mean nothing historical performance is the only thing you can rely upon. Burstable infrastructure can be a way to survive spikes of usage. No surprise here, high level presentation explaining why Amazon did this not some much how they did it. Massive growth in usage, several billion “objects” now stored in S3, replicated across at least two locations for each object.

7. Microsoft Research
As usual very cool presentations of advanced technology research:
Collaboration - education focused, on distributed meetings, multi camera panoramic views etc.
Cool use of surface computing to show other peoples remote actions while playing chess remotely (See CSI Miami if you have not seen this in action)
Natural language processing, mindnet - foundation for a Universal Translator – this looks like some real progress, they showed some impressive stats about the performance of automated translation compared with human translations of knowledge base articles.
Drag and pop - slick predictive destination for touch screens, predicts where your finger is going and then brings the destination to you! Awesome for multi monitor scenarios especially on screens with bezels.

8. Myspace
Very interesting deep dive into the underpinnings of one of the most active web sites on the globe.
Disk IO limits way before disk capacity, new large disks don’t help as they don’t spin any faster. Data centered paired up regionally to provide DR via 2.5 GBS connections. Backups are just dealt with by replication out of region. Backups are after all just another copy to another medium, everything is live forever there are no offline backups any more.
Billions of static objects distributed around their distributed file systems. Active active active datacenters fully redundant persisting transactions across three to ensure DR and HA. Any single data center can sustain a complete failure without affecting the user. Communications links too latent for applications to be unaware of the latency, build this into your transactional model of publishing.
Interesting take of porn scanning, they use humans to scan images and content and add checksums of blacklisted content to a list to prevent future uploads of the content.

9. Paul Glenn - Leading Geeks
Second time I have heard Paul speak, excellent speaker on Geek Leadership. Took many new ideas away, read his book if you want to get them all. Here are just a few that stuck in my mind:
Select Wisely - Pick people who are motivated to do your project. Halo effect – good people on your team will be a great recruitment mechanism. You can’t motivate geeks, all you can do is de-motivate them, so tread lightly to do the least damage.
Fastest way to make a project fail, tell the geeks failure is not an option….