Requirements for Fault Tolerance in SIGTRAN

1
Requirements for Fault Tolerance in SIGTRAN

• Transparent support of server pooling (2n, nk,
  etc)
• Automatic fail-over to alternative server
  (controllable by application)
• Support plug-and-play type of service/server
  addition or subtraction (i.e, dynamic
  scalability)

2
Requirements (cont.)

• Support load balancing within a server pool
• Individual server instance addressable in a
  server pool
• Simple programming interface for fast application
  development

3
Model and Goals

• Push down as much of fault tolerance issues as
  possible to hide them from the application.
• Divide fault tolerance into two layers - link
  level (w/ SCTP) and application level (w/ DDP)
• Name-based addressing of processes.

4
Model and Goals (cont.)

• Applications are dynamic - names enter and exit
  the name space frequently.
• Flexible name syntax can be either ASCII or
  binary.
• Name space must be totally distributed and fault
  tolerant, i.e, not a single point of failure
  itself.