On the possibility and impossibility of achieving clock synchronization

Untitled

“…This forces C p (t) to be within a linear envelope of real-time, thereby excluding "degenerated" cases like C p (t) ≡ 0, (see [DHS86]). …”

“…It will allow us to exploit the fact that masking f symmetric faults with OA requires only n ≥ 2f + 1, whereas n ≥ 3f + 1 is needed if all faults are asymmetric ones ( [DHS86]). Since a large number of asymmetric faults is very unlikely in practice, see [Sch95], this effectively leads to a smaller n for tolerating a given number of faults, see (3).…”

Consul: a communication substrate for fault-tolerant distributed programs

“…Examples of such services include broadcast services-to deliver messages to the cooperating processes reliably and in some consistent order, membership services-to maintain a consistent" system-wide view of which processes are functioning and which have failed, and recovery services-to recover a failed prece~s. Much work has been done on developing algorithms and implementation for these services [Birm87,Chan84,Cris88,Dole84,Ko087,Lamp78]. Chapter 2 elaborates on the characteristics of each of these.…”

“…However, clocks may be synchronized in presence of any number of Byzantine failures if an authentication scheme is used [Dole84]. Optimal algorithms for clock synchronization under different failure scenarios are also known [Srik87].…”

Clock Synchronization in the Byzantine-Recovery Failure Model

“…There is considerable literature devoted to the design and implementation of clock synchronization algorithms; see [1][2][3][4][5] for an overview. Some algorithms are specified for environments in which processes may crash [6], may suffer timing failures [7], or may execute arbitrarily bad operations [1,[8][9][10].…”