Which of the following is false about the Repulsion forces approach in large scale coordinator election
each token is “repulsed” by nearby tokens (token holder sends token to another peer if too many tokens nearby)
token must be held by a node for a set time period before node can become superpeer
each node holding a token learns about other tokenholders
a superpeer which receives another token sends it to the lowest prime numbered neighbour
How can a coordinator failure be addressed in a centralised mutual exclusion system?
By using election algorithms to change the coordinator
By using triple modular redundancy to compare the results of two coordinators
By emitting a heartbeat that informs other nodes when the coordinator cannot provide a lock
By moving the coordinator periodically to the next node in the token ring
Which of these is not a drawback of a typical distributed mutual exclusion algorithm?
large number of messages required
one slow process slows down others
more points of failure
locks may be granted unfairly
Which of the following is true about Lamport's algorithm
when a message is received, its time is compared against the local clock. If the local clock is less than C(b), it is set to C(b) + 1
messages carry their sending time according to the receiver's clock e.g., C(b)
between any two events, the clock must tick at least twice (i.e. one on the transmitting side, one on the receiving side)
Events occurring in processes that do not interact (even indirectly through third parties) are said to be concurrent
In the bully algorithm with nodes 1,2,3,4,5,6,7 if 7 is the coordinator and 4 notices the coordinator has crashed, who will be the new coordinator?
6
4
1
7 after it reboots
Which of the following is false about wireless network clock synchronisation algorithms
Clock synchronisation is difficult because nodes cannot always contact one another
Nodes are resource-constrained
Receivers record difference between reference message timestamp and their own clock
The coordinating node must transmit at a higher power level to ensure all nodes can receive clock broadcasts
Which of the following is false about Lamport's algorithm?
if a and b are events in the same process, and a occurs before b, then a - > b is true
if a is the event of a message being sent by one process, and b is the receipt of that message in another process, then a ->b is also true
a -> b and b -> c, then a -> c
if a is the event of a message being sent by one process, and b is the receipt of that message in another process, then b -> a is also true
Which of these are clock synchronisation protocols
Network Time Protocol, The Berkeley Algorithm
Network Time Protocol, Town Clock Algorithm, The Berkeley Algorithm
UTC lock algorithm, NTP
Network Time Protocol, Town Clock Algorithm
Which is true about NTP?
Stands for Numerical Time Protocol
Daemon calculates new time based on average of reported values
In general, synchronising with stratum-k server makes you stratum-k+1
Stratum-1 servers typically have direct access to a reference clock
What issue is addressed by a vector clock?
Lamport’s algorithm does not capture causality
Sometimes messages are received out of order
Sometimes messages are created at precisely the same time
Sometimes events must be recorded before an event which occurred after it
Which is true about the Berkeley algorithm?
Uses averaging approach to correct clocks
Daemon tells each node how to adjust its clock
Servers are divided into strata reflecting the accuracy of their clocks
Lower stratum servers are assumed to be more accurate => A will try to synchronise with B’s clock
Which of the following is false about superpeers
Large-scale systems may require many local coordinators, creating the need for superpeers
Should offer low-latency access to other superpeers
Serve no more than a set number of regular nodes
Can be selected by reserving a fraction of the identifier space in a distributed hash table containing all nodes
In the ring election algorithm with nodes 1,2,3,4,5,6,7 if 7 is the coordinator and 5 and 2 both notice which node becomes the new coordinator?
5
5 or 2
2
In a token ring mutual exclusion algorithm, how does a node respond if it currently holds the resource?
The token is passed on to the next node in the ring when it is finished accessing the shared resource
It sends a message in the opposite direction back to the requester stating that it holds the lock
It sends a message in the forward direction to the requester stating that it holds the lock
The previous node sends a message back to the requester when it doesn't get a response from the node holding the resource
In a regular distributed mutual exclusion system, how does a node reply if it currently holds the resource?
It ignores the message
It sends a NACK response
It queues the message until it is finished with the resource then replies OK
It replies to the next node in the token ring
Which of these is not an advantage of centralised mutual exclusion?
easy to implement
low message overhead
fair (access requests are processed in order)
improves performance with high message volumes
