22-03-2011, 10:21 AM
[attachment=10707]
RAIN Platform
Heterogeneous network of nodes and switches
Proof of Concept: Video Server
Video client & server on every node.
Limited Storage
Insufficient storage to replicate all the data on each node.
Encoding
Encode video using 2-of-4 code.
Decoding
Retrieve data and decode.
Node Failure
Dynamically switch to another node.
Link Failure
Dynamically switch to another network path
Switch Failure
Dynamically switch to another network path.
Node Recovery
Continuous reconfiguration (e.g., load-balancing).
Features
High availability:
• tolerates multiple node/link/switch failures
• no single point of failure
Efficient use of resources
• multiple data paths
• redundant storage
• graceful degradation
RAIN Project: Goals
Efficient, reliable distributed computing and storage systems
Interconnect Topologies
Goal:
lose at most a constant number of nodes for given network loss
Resistance to Partitions
Large partitions problematic for distributed services/computation
Related Work
Embedding hypercubes, rings, meshes, trees in fault-tolerant networks:
Bus-based networks which are resistant to partitioning:
RAIN Platform
Heterogeneous network of nodes and switches
Proof of Concept: Video Server
Video client & server on every node.
Limited Storage
Insufficient storage to replicate all the data on each node.
Encoding
Encode video using 2-of-4 code.
Decoding
Retrieve data and decode.
Node Failure
Dynamically switch to another node.
Link Failure
Dynamically switch to another network path
Switch Failure
Dynamically switch to another network path.
Node Recovery
Continuous reconfiguration (e.g., load-balancing).
Features
High availability:
• tolerates multiple node/link/switch failures
• no single point of failure
Efficient use of resources
• multiple data paths
• redundant storage
• graceful degradation
RAIN Project: Goals
Efficient, reliable distributed computing and storage systems
Interconnect Topologies
Goal:
lose at most a constant number of nodes for given network loss
Resistance to Partitions
Large partitions problematic for distributed services/computation
Related Work
Embedding hypercubes, rings, meshes, trees in fault-tolerant networks:
Bus-based networks which are resistant to partitioning: