This is a great conference. It’s great to be back in NYC. There is no place like NYC. The sights. The smells. Seeing my family. All great. Here is a sample of today’s sessions …
Keynotes:
Tim O’Reilly: Web 2.0 is a data operating system; Work on stuff that matters; Create more value than you capture; Great potential in big problems; Vote: don’t let those who don’t participate pick our leadership.
Clay Shirkey: Information OVERLOAD. New problem? Not really. Soon after Gutenberg revolutionized printing there were too many books in print for most people to read in a lifetime. This problem has been with us for about 500 years. We need a new perspective on data flow. We need better filters.
Cloud Panel ::
An interesting discussion about what the cloud is. And isn’t. Best bits? Remember to consider scaling down when you are obsessed with scaling up or out. It’s always good to be as close to your end-users as feasible. When cpu’s are sold as a utility, coding practices will change lower utilization costs.
Digg Scaling :: Joe Stump
Stump Dump:
Scaling can cause severe hair loss.
Your mother lied: Share nothing. Share nothing architectures are the key to scaling out.
Decentralize; expect failure; just add boxes.
Cache forever; explicitly expire; develop a chain of responsibility.
Partition your data!
Joe Stump. Awesome.
Sequel to SQL
Whoa. As a (relational) database guy this talk was fascinating and scary. Relational databases don’t work in the cloud. Period. Geir explored plate spinning on EC2, Google’s BigTable, Amazon’s SimpleDB and 10gen’s Mongo data persistence platforms. It’s a different world now. “Eventually consistent” is my new, favorite term. Bottom line is we need to think about this problem from a new perspective and develop new solutions.
Scaling Meebo : Sandy Jen
The only women geek presenter so far. Sandy did a great presentation. Ironically, in my mind she was the geekiest as Meebo is a C/C++ application. She cautioned us to carefully decide when to be synchronous and when to be asynchronous. She confirmed a personal truism for me: Nothing simulates real life like real life. Load testing can provide valuable insights, but you don’t know if it’s going to work until you let your application loose. Also, look at alternative to Apache. lighttpd rocked their world.
Alix Iskold :: Amazon Web Services
They rock. Why?
- Pay per use model
- Instant scalability
- Reliable/Redundant/Secure
- Simple REST/SOAP API
- Amazon’s Experience and Commitment (overlooked and unappreciated)
A great talk about his experience running
AdaptiveBlue as a startup on Amazon AWS.
Wine 2.0
Wine meets the web. Really. Check out
snooth. Cool guys who have a wine api. I want to find a reason to integrate with this platform. I tasted some good wine. I talked to a few interesting people. I thought about wine and the internet.