@techreport{TR-IC-10-13,
  number = {IC-10-13},
  author = {Gustavo Maciel Dias Vieira and Islene Calciolari Garcia
                  and Luiz Eduardo Buzato},
  title = {{Seamless Paxos Coordinators}},
  month = {April},
  year = {2010}, 
  institution = {Institute of Computing, University of Campinas},
  note = {In English, 18 pages.
    \par\selectlanguage{english}\textbf{Abstract}
  In  Paxos, failures  can cause  the replacement  of  the coordinator
  agent, a  key process of this consensus  algorithm.  The replacement
  of the coordinator, in its turn, leads to a temporary unavailability
  of  the  application  implemented  atop  Paxos.   Solutions  to  the
  unavailability  problem  have  been  sought because  of  the  widely
  recognized utility  of Paxos as  a building block  of fault-tolerant
  distributed applications.  So far,  the problem has been lessened by
  reducing  the  coordinator replacement  rate  through stable  leader
  election. We  have observed that  the recovery of the  newly elected
  coordinator's state  is at the  core of the  unavailability problem.
  Thus, in  this paper we present  a new solution to  the problem that
  allows the  state recovery to occur concurrently  with new consensus
  rounds.  The complexity and  performance of the solution is assessed
  both  theoretically and  experimentally.  Experimental  results show
  that the unavailability period  is removed during the replacement of
  the coordinator making it seamless to the application.
  }
}