Resilience Flashcards
What is the definition of node connectivity of a graph?
the minimum number of links (nodes) whose
removal disconnects G
What is Bhandari’s algorithms used for?
Find k different paths between the src and dst without the paths sharing common links.
Explain how Bhandari’s algorithm works.
Might have to solve it on the exam.
You first find the minimum weighted path between src and dst. Then you take this P1 and flip the directions of the links as well as revers the weights. No you again try to find the minimum weighted path (P2). All the interlacing links can now be discarded, such that two (different) paths still remain. Then you reverse all used links and weights and see if another path is still viable, etc.
What is the difference between Maximally disjoint paths and bhandari’s algorithm?
In max dis path you try to find the paths with minimal interlacing links, bhandari only finds paths with no interlacing links.
What is SRG-disjoint paths? (explain)
Shared risk group disjont paths: Find k paths from src to dst, such that the paths share no common risk groups.
What is ATTR and what does this value represent?
the Average Two-Terminal Reliability:
The number of connected node pairs in a graph / total number of node pairs in the graph
What are the advantages of SDN failure Restoration approach?
– Straightforward
– Only uses computation time after failure
– Does not require extra network resources
What are the disadvantages of SDN failure Restoration approach?
– Very slow
– Lot of control packets/processing time at once
What are the disadvantages of SDN failure Restoration approach?
– Very slow
– Lot of control packets/processing time at once
What kind of protection measures can be done in SDN to aid resilliance?
- Pre-install backup paths/trees
- Quick switchover after failure
- Fast Built-in protection mechanism: Fast Failover
How does Fast Failover Group work?
Each switch maintains a list of buckets for each flow. Each bucket contains a** liveness watch port and an action**. When a port goes down, the switch will quickly select the next bucket in order to maintain that flow. (often also includes a tag on that packet to notify others of the new flow).
What is the benefit of Fast Failover groups?
No input of the controller required; no new flows need to be installed on multiple switches during every link/node failure. This is guaranteed to be much quicker.
What is downside of Fast Failover compared to controller restoration?
– Backups need to be computed and installed for every possible link/node failure
– Backup paths can be worse than in restoration
– Group tables are in limited supply
