Deadlock detection in distributed systems

RezaRamezani68 7,269 views 20 slides Sep 23, 2013
Slide 1
Slide 1 of 20
Slide 1
1
Slide 2
2
Slide 3
3
Slide 4
4
Slide 5
5
Slide 6
6
Slide 7
7
Slide 8
8
Slide 9
9
Slide 10
10
Slide 11
11
Slide 12
12
Slide 13
13
Slide 14
14
Slide 15
15
Slide 16
16
Slide 17
17
Slide 18
18
Slide 19
19
Slide 20
20

About This Presentation

No description available for this slideshow.


Slide Content

Deadlock Detection in Distributed Systems Dr. Davarpanah Research: Kaywan Zayer Reza Ramezani Mohammad Khosravi

Deadlock Introduction Strategies to handle deadlock Ostrich algorithm (ignore it) Detection (let occur, detect, recover) Prevention (make it impossible to o ccur) Avoidance (careful resource allocation)

Deadlock Detection Centralized Single point of failure bottleneck Distributed Soojung Lee Monjurul Alom et al Farajzadeh et al

Introduction Process dependency wait-for graph (WFG) Request model AND OR p out-of q Communication model Just message communication Bounded delay, reliable protocol

Soojung Lee Dependency Model Reduced wait-for graph (RWFG) knot Request model OR Blocked process can participate in deadlock detection algorithm

Message Probe Distributed Spanning Tree (DST) Path-string decoding Unique id Reply ACTIVE REPORT

Example (i) WFG (ii) DST

Example (iii) RWFG after receive ACTIVE (iv) RWFG after receive REPORT (v) Final RWFG

Monjurul Alom et al Distributed Database System Dependency model Transaction wait-for graph (TWFG) Request model AND Linear Transaction Structure (LTS) Distributed Transaction Structure (DTS)

Algorithm Scheme • LTS for each local site • DTS for global resource transaction communication • Priority Id for each transaction in each of the sites • The local and global cycles • Abort of the victim transaction based on the cycles

Example TWFG

Example LTS2 : {3->5, 5->6, 6->3} {7->10, 10->9, 9 -> 7 } {10->9, 9->11, 11->10} LTS3 :

Example

Example DTS1 : {4-> 3 , 3 -> 3 , 3 ->4, 4->4} DTS2 : {6-> 4 , 4 -> 8 , 8 ->7, 7->6}

Detected deadlock, breaking it

Final TWFG

Farajzadeh et al History-based edge chasing Probe message Request model p out-of q Messages Probe Clean-up Ok Deny

Algorithm Receive probe In memory and id = received id Prefix of received route-string deadlock Otherwise Save probe with smaller path-string and propagate received probe Deadlock Initiator Clean-up and suicide Otherwise Request permission for suicide

Comparison First an third algorithm Both use probe messages Build WFG Different in message number, delay, message size and resolution scheme Sujung lee farajzadeh Number of message 2e <=e Message size d+2 <=O(d) delay O( dlogn ) n resolutaion yes Yes, 2 message e : number of edge in WFG, n : number of node, d : diameter of WFG
Tags