Pesadelo de SDIS
6 - Consider Chord, with a hash function whose result has m bits. If all statements are true, select: “All statements are true”, otherwise select the false statement.
[?] Key resolution requires O(m) steps
[ ] Each node keeps routing information on at most m nodes, not all. (each Chord node maintains a finger table of at most m entries)
[ ] The finger table of a node has more entries on nodes closer to it (in the ring) than entries on nodes further away. (true, normalmente ele sabe o successor, o n+2 e n+4)
[?] All true.
Dúvida: Segundo eles O(log(N)) = O(log(2^m)) mas N pode ser != de 2^m logo não sei se faz sentido...
14 - Consider the Switch Tree Protocol. Choose the true statement.
[?] The one-hop protocol, as presented in class, does not tolerate the failure of the root node.
[ ] In the absence of faults, the switch-sibling protocol does not allow the level of a node to decrease.
[ ] The two-hop protocol allows topology changes that are not possible with the one-hop protocol, even after multiple topology changes.
[ ] All false.
Dúvida: Não diz em lado nenhum sobre a root
16 - Consider view-synchronous multicast. If all statements are true select “All statements are true”. Otherwise, select the false statement.
[?] Only correct processes in a view, deliver the same set of messages in that view..
[ ] All correct processes in a view, always deliver the same set of messages in that view.
[ ] All processes that move from one view to the next, always deliver the same set of messages in the first view.
[?] All true.
Dúvida: Aquele only na primeira... não dizem nada sobre processes que não são correct nos ppt
19 - Consider concurrency at the processing elements of a distributed system. If all statements are true select “All statements are true”. Otherwise, select the false statement.
[ ] Multi-threaded designs facilitate real parallelism in a multi-core platforms.
[?] The number of threads is the bottleneck of pure multi-threaded-based designs.
[ ] Pure event based designs may be better than pure multi-threaded-based designs on “uni-processor” platforms.
[?] All true.
Dúvida: Diogo Rodrigues diz que o bottleneck depende do paralelismo suportado pelo CPU e não do número de threads mas não sei se concordo.
22 - Consider Remote Procedure Call (RPC) and parameter passing by copy-and-restore. Choose the true statement.
[ ] For efficiency reasons, most RPC systems pass parameters by copy-and-restore.
[ ] The copy-and-restore parameter passing works only with tree-like structures, i.e. when each object is pointed by at most another object.
[?] RPC systems that use copy and restores parameter passing cannot use addresses.
[?] All false.
Dúvida: ??