Slides for Chapter 15: Coordination and Agreement

Size: px

Start display at page:

Download "Slides for Chapter 15: Coordination and Agreement"

Brian Parsons
6 years ago
Views:

1 Slides for Chater 15: Coordination and Agreement From Coulouris, Dollimore, Kindberg and Blair Distributed Systems: Concets and Design Edition 5, Addison-Wesley 2012

2 Overview of Chater Introduction Distributed mutual exclusion Elections Coordination and agreement in grou communication (ski) Consensus and related roblems (ski) 2

3 Introduction Covers two areas: Coordinating actions in a distributed system Distributed rocesses agreeing on a result value Assumes reliable communication channels for simlicity (failure is masked by a reliable communication rotocol) Detecting that a rocess has failed can be reliable or unreliable Use timouts Unreliable: relies unsusected or susected Reliable: relies unsusected or failed 3

4 Figure 15.1 A network artition Cras hed router Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

5 Overview of Chater Introduction Distributed mutual exclusion Elections Coordination and agreement in grou communication Consensus and related roblems 5

6 Distributed mutual exclusion Known as critical section roblem: Only one rocess can be in critical section the rocess with the token that allows access to the resource Oerations include the following: enter() requests access; can be granted or blocked resourceaccess() access resource in critical section exit() leave critical section other rocesses may now enter Conditions: ME1 (safety) at most one rocess in critical section ME2 (liveness) requests eventually succeed ME3 (ordering) requests follow haened-before relationshi 6

7 Distributed mutual exclusion Various algorithms: Central server algorithm Ring-based algorithm Algorithm using multicast and logical clocks Voting algorithm Others 7

8 Figure 15.2 Server managing a mutual exclusion token for a set of rocesses Queue of requests Server Grant token 1 1. Request token 2. Release token Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

9 Figure 15.3 A ring of rocesses transferring a mutual exclusion token 1 2 n 3 4 Token Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

10 Figure 15.4 Ricart and Agrawala s algorithm On initialization state := RELEASED; To enter the section state := WANTED; Multicast request to all rocesses; request rocessing deferred here T := request s timestam; Wait until (number of relies received = (N 1)); state := HELD; On receit of a request <T i, i > at j (i j) if (state = HELD or (state = WANTED and (T, j ) < (T i, i ))) then queue request from i without relying; else rely immediately to i ; end if To exit the critical section state := RELEASED; rely to any queued requests; Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

11 Figure 15.5 Multicast synchronization Rely Rely 34 Rely 2 34 Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

12 Figure 15.6 Maekawa s algorithm art 1 On initialization state := RELEASED; voted := FALSE; For i to enter the critical section state := WANTED; Multicast request to all rocesses in V i ; Wait until (number of relies received = K); state := HELD; On receit of a request from i at j if (state = HELD or voted = TRUE) then queue request from i without relying; else send rely to i ; voted := TRUE; end if For i to exit the critical section state := RELEASED; Multicast release to all rocesses in V i ; On receit of a release from i at j if (queue of requests is non-emty) then remove head of queue from k, say; send rely to k ; voted := TRUE; else voted := FALSE; end if Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

13 Overview of Chater Introduction Distributed mutual exclusion Elections Coordination and agreement in grou communication Consensus and related roblems 13

14 Elections Election algorithms: Used to choose a articular rocess for a role for examle, to choose a central server for distributed algorithms that require a central server Process can call an election; for examle, if it detects that central server has failed Requirements: E1: (safety) only one rocess is elected each articiant sets electedi either to P or to undefined if it does not know the elected rocess yet (P will be the articiant with the largest rocess id) E2: (liveness) all rocesses articiate and either set electedi to the elected rocess or crash 14

15 Elections Election algorithms: Ring-based algorithm Bully algorithm 15

16 Figure 15.7 A ring-based election in rogress Note: The election was started by rocess 17. The highest rocess identifier encountered so far is 24. Particiant rocesses are shown in a darker colour Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

17 Figure 15.8 The bully algorithm The election of coordinator 2, after the failure of 4 and then 3 Stage 1 1 election answer election 2 answer 3 election C 4 Stage 2 election answer election C 4 timeout Stage Eventually... coordinator C Stage Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

18 Overview of Chater Introduction Distributed mutual exclusion Elections Coordination and agreement in grou communication (ski) Consensus and related roblems 18

19 Figure 15.9 Reliable multicast algorithm Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

20 Figure The hold-back queue for arriving multicast messages Message roc es sing deliv er Hold-bac k queue Deliv ery queue Incoming mes sages When deliv ery guarantees are met Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

21 Figure Total, FIFO and causal ordering of multicast messages Notice the consistent ordering of totally ordered messages T 1 and T 2, the FIFO-related messages F 1 and F 2 and the causally related messages C 1 and C 3 and the otherwise arbitrary delivery ordering of messages. T 1 F 1 F 2 T 2 F 3 Time C 1 C 2 C 3 P 1 P 2 P 3 Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

22 Figure Dislay from bulletin board rogram Bulletin board: os.interesting Item From Subject 23 A.Hanlon Mach 24 G.Joseh Microkernels 25 A.Hanlon Re: Microkernels 26 T.L Heureux RPC erformance 27 M.Walker Re: Mach end Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

23 Figure Total ordering using a sequencer Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

24 Figure The ISIS algorithm for total ordering P Message P Agreed Seq 2 3 P 1 P 3 Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

25 Figure Causal ordering using vector timestams Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

26 Figure Consensus for three rocesses P 1 d 1 :=roceed d 2 :=roceed P 2 v 1 =roceed 1 v 2 =roceed Consens us algorithm v 3 =abort P 3 (c rashes ) Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

27 Figure Consensus in a synchronous system Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

28 Figure Three Byzantine generals 1 (Commander) 1 (Commander) 1:v 1:v 1:w 1:x 2:1:v 2 3 3:1:u 2:1:w 2 3 3:1:x Faulty rocesses are shown coloured Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

29 Figure Four Byzantine generals 1 (Commander) 1 (Commander) 1:v 2:1:v 1:v 1:v 2 3:1:u 3 1:u 2:1:u 1:w 1:v 2 3:1:w 3 4:1:v 4:1:v 4:1:v 4:1:v 2:1:v 3:1:w 2:1:u 3:1:w 4 Faulty rocesses are shown coloured 4 Instructor s Guide for Coulouris, Dollimore, Kindberg and Blair, Distributed Systems: Concets and Design Edn. 5

Distributed Systems (5DV147)

Distributed Systems (5DV147) Mutual Exclusion and Elections Fall 2013 1 Processes often need to coordinate their actions Which rocess gets to access a shared resource? Has the master crashed? Elect a new