Principles and Practices of Interconnection Networks

CHAPTER 21

Error Control

Many applications of interconnection networks require high reliability and avail-

ability. A large parallel computer requires that its interconnection network operate

without packet loss for ten thousands of hours. An Internet router can accept a small

amount of packet loss, but the router itself must remain up with an availability of

99.999% — five minutes of downtime allowed per year. I/O systems have similar

availability requirements.

Interconnection networks are often composed of hundreds (or thousands) of

components — routers, channels, and connectors — that collectively have failure

rates higher than is acceptable for the ...