Views on realizing distributed systems:
- Integrative view
- Expansive view
Integrative view:
Connecting existing systems into a larger one
Expansive view:
Extending an existing system with additional computer
Decentralized system:
A system where resources are necessarily spread across computers
Distributed system:
A system where resources are sufficiently spread across computers
Perspectives on distributed systems:
- Architecture
- Process
- Communication
- Coordination
- Naming
- Consistency and replication
- Fault tolerance
- Security
Common misconceptions about centralized solutions:
- They don’t scale
- They have a single point of failure
Distributed system design goals:
- Sharing resources
- Transparency
- Openness
- Scalability
Middleware layer:
A transparency technique that puts a layer between applications and the OS
Dependability trust:
An entity can be assured that another will perform up to its expectations
Scale in distributed system components:
- Size scalability
- Geographical scalability
- Administrative scalability
Administrative scalability:
Number of administrative domains
Problems with geographical scalability:
- Cannot go from LAN to WAN
- WANs are unreliable
- Lack of multipoint communication
Size scalability:
Number of users or processes
Geographical scalability:
Maximum distance between nodes
Cluster computing:
High-end homogeneous systems connected through a LAN
Techniques for scaling:
- Partition data across machines
- Replicate and cache data at different machines
Grid computing:
Heterogeneous nodes from everywhere connected through a WAN
Virtual organization:
A grouping of users for resource allocation
Distributed pervasive systems:
Distributed systems in which nodes are small, mobile, and embedded in a larger system that naturally blends into its environment
Ubiquitous:
Continuously present
Distributed pervasive system subtypes:
- Ubiquitous systems
- Mobile computing
- Sensor networks
Ubiquitous system core elements:
- Distribution
- Interaction
- Context awareness
- Autonomy
- Intelligence
Sensor network node characteristics:
- Many
- Simple
- Battery-powered or battery-less
