Four major characteristics of a 21st century ar...

Vogels contends that customers can now build the systems they have always wanted, but were prevented from doing so in the past. He said that factors like capital, capacity,  geography, physics, people, and scope are no longer constraints to building secure, scalable, fault tolerant, high performing and cost effective applications. Vogels sees everything as a programmable resource: data centers, networks, compute, storage, databases and load balancers. A 21st century architecture embraces automation as a way to increase efficiency and decrease cost. Vogels pointed out four major characteristics of a 21st century architecture.

• Controllable. New systems should be decomposed into small, loosely coupled, stateless building blocks. Software should be comprised of individual units that you can control. These units may pivot on scale, fault tolerance or other critical dimensions. Vogels used an example from IMDB where an initial integration architecture created a tight linkage that forced IMDB to scale whenever Amazon.com did. Instead, they decomposed and decoupled the systems further by using S3 as an intermediary.  This allowed each layer to scale independent of the other.
• Resilient. Vogels says that the first priority of any software is to protect the customer. Any sensitive data should be encrypted at rest. As an example, Amazon.com encryptseverything at rest and in transit. Production system should be geographically distributed in order to survive failure in a single data center. Vogels reminded the audience that failure is always around the corner, and he encouraged architects to not think of failures as exceptions, but as another form of deployment.
• Adaptive. Modern software shouldn’t be dependent on fixed resources. Adapt to different circumstances, but don’t be constrained by them. As an example, S3 was originally scoped to hold 20 billion objects, and quickly had to be refactored to address the explosive growth that now has the service with over 1 trillion objects. All of this was done without disrupting the existing customer base. Vogels told the crowd to assume nothing and avoid applying constraints up front that will prevent future changes.
• Be Data Driven. Instrument everything. Don’t rely on predictions, but use reality and data to make decisions. If you don’t collect the data, reminded Vogels, you cannot act upon it. Network services like the AWS Elastic Load Balancer can make decisions based on more than just low-level system metrics, but also business-level data points. Vogels encouraged architects and operational staff to not be shy about collecting data and take advantage of tools like the new Data Pipeline to aggregate the data for analysis.