Survivor: an Enhanced Controller Placement Strategy for Improving - - PowerPoint PPT Presentation

▶
survivor an enhanced controller placement strategy for
SMART_READER_LITE
LIVE PREVIEW

Survivor: an Enhanced Controller Placement Strategy for Improving - - PowerPoint PPT Presentation

Sep 21, 2022 102 likes •434 views

Survivor: an Enhanced Controller Placement Strategy for Improving SDN Survivability Lucas F. Mller , Rodrigo R. Oliveira, Marcelo C. Luizelli, Luciano P. Gaspary, Marinho P. Barcellos Federal University of Rio Grande do Sul (UFRGS), Brazil

slide-1
SLIDE 1

Survivor: an Enhanced Controller Placement Strategy for Improving SDN Survivability

Lucas F. Müller, Rodrigo R. Oliveira, Marcelo C. Luizelli, Luciano P. Gaspary, Marinho P. Barcellos Federal University of Rio Grande do Sul (UFRGS), Brazil 57th IEEE Global Communications Conference (GLOBECOM 2014) December 8 - 12, 2014 Austin, Texas – EUA

slide-2
SLIDE 2

Lucas F. Müller Survivor: Controller Placement Survivability – 2

Software-Defined Networking Design

  • Changing the way networks are designed and managed
  • Separates the control plane from the data plane
  • Moves the control logic to an external entity (Controller)
  • Controller provides resources and abstractions to facilitate

programming … Despite its benefits, SDN created an inherent dependency relationship between forwarding devices and the controller.

slide-3
SLIDE 3

Lucas F. Müller Survivor: Controller Placement Survivability – 3

Software-Defined Networking Design

Forwarding devices Set of SDN Controllers Controller Placement Problem

slide-4
SLIDE 4

Lucas F. Müller Survivor: Controller Placement Survivability – 4

Software-Defined Networking Design

Controller Placement Problem

slide-5
SLIDE 5

Lucas F. Müller Survivor: Controller Placement Survivability – 5

Software-Defined Networking Design

Controller Placement Problem

slide-6
SLIDE 6

Lucas F. Müller Survivor: Controller Placement Survivability – 6

Software-Defined Networking Design

Controller Placement Problem

slide-7
SLIDE 7

Lucas F. Müller Survivor: Controller Placement Survivability – 7

Software-Defined Networking Design

Controller Placement Problem

slide-8
SLIDE 8

Lucas F. Müller Survivor: Controller Placement Survivability – 8

Software-Defined Networking Design

Controller Placement Problem

slide-9
SLIDE 9

Lucas F. Müller Survivor: Controller Placement Survivability – 9

Software-Defined Networking Design

Controller Placement Problem

slide-10
SLIDE 10

Lucas F. Müller Survivor: Controller Placement Survivability – 10

Software-Defined Networking Design

1 1 1 1

Controller Placement Problem

slide-11
SLIDE 11

Lucas F. Müller Survivor: Controller Placement Survivability – 11

Software-Defined Networking Design

1 1 1 1 2 2 2 2

Controller Placement Problem

slide-12
SLIDE 12

Lucas F. Müller Survivor: Controller Placement Survivability – 12

Software-Defined Networking Design

1 1 1 1 2 2 2 2 3 3 3 3

Controller Placement Problem

slide-13
SLIDE 13

Lucas F. Müller Survivor: Controller Placement Survivability – 13

Software-Defined Networking Design

1 1 1 1 2 2 2 2 3 4 4 3 4 3 4 4 3 4 4 4 4

Ok, the control plane design is ready. Controller Placement Problem

slide-14
SLIDE 14

Lucas F. Müller Survivor: Controller Placement Survivability – 14

“The network is down.”

slide-15
SLIDE 15

Lucas F. Müller Survivor: Controller Placement Survivability – 15

Software-Defined Networking Design

1 1 1 1 2 2 2 2 3 4 4 3 4 3 4 4 3 4 4 4 4

X

1 1 1 1 2 2 2 2 3 4 4 3 ? 3 ? ? ? ? ? ? ?

X X

1 1 1 1 2 2 2 2 3 4 4 3 4 3 4 4 3 4 4 4 4

# 5 # 4 # 4 # 4 # 9

Single link failures Multiple connectivity failures Controller overload

slide-16
SLIDE 16

Lucas F. Müller Survivor: Controller Placement Survivability – 16

Controller Placement Strategy for Improving SDN Survivability

Goal: novel controller placement strategy that deals with control plane survivability in large scale SDN networks. Provide and maintain network services in face of

  • perational challenges

React and attempt to recover from harmful events

slide-17
SLIDE 17

Lucas F. Müller Survivor: Controller Placement Survivability – 17

Outline

  • Introduction: context and motivation
  • Proposed Approach: strategy and modeling
  • Results: resilience and overload
  • Conclusion
slide-18
SLIDE 18

Lucas F. Müller Survivor: Controller Placement Survivability – 18

Proposed Approach

Goals

– Connectivity

Increase path diversity between device-controller

– Capacity

Avoid controller overload

– Recovery

Define a methodology for composing smarter failover mechanisms

slide-19
SLIDE 19

Lucas F. Müller Survivor: Controller Placement Survivability – 19

Proposed Approach: Overview

Divided in two complementary parts

– Defines the placement of controllers instances – Compose the list of backup controllers for each device in the network

slide-20
SLIDE 20

Lucas F. Müller Survivor: Controller Placement Survivability – 20

Proposed Approach: two complementary parts

– Defines placement for controller instances

32 33 34 1 6 24 25 7 4 5 8 9 28 18 15 16 17 14 12 20 13 11 21 26 22 27 23 29 10 19 3 30 39 38 37 36 2 35 31 3 3 2 5 3 2 2 2 2 4 4 4 4 5 5 4 1 4 1 1 1 1 1 1 4 2 5 5 5 2 5 3 3

Network topology Controller placement

slide-21
SLIDE 21

Lucas F. Müller Survivor: Controller Placement Survivability – 21

Proposed Approach: two complementary parts

– Specifies backup controllers for each device in the network

3 3 2 5 3 2 2 2 2 4 4 4 4 5 5 4 1 4 1 1 1 1 1 1 4 2 5 5 5 2 5 3 3

2 4 5 1 4 2 1 3 4 1 2 5

slide-22
SLIDE 22

Lucas F. Müller Survivor: Controller Placement Survivability – 22

Proposed Approach: modeling

Optimal Linear Model for Controller Placement

– Strategy modeled as optimization problem – Achieve the optimal solution – Survivor strategy: Integer Linear Program, 1 objective (maximize connectivity between device-controller)

Heuristics for Defining Lists of Backup Controllers

– Compose the lists of backup controllers – Eliminating the need to manually determine the list – Proximity and Residual capacity-based heuristics – Proposed generic framework for designing heuristics

slide-23
SLIDE 23

Lucas F. Müller Survivor: Controller Placement Survivability – 23

Outline

  • Introduction: context and motivation
  • Proposed Approach: strategy and modeling
  • Results: resilience and overload
  • Conclusion
slide-24
SLIDE 24

Lucas F. Müller Survivor: Controller Placement Survivability – 24

Methodology

Configuration

– Three different WAN topologies: Internet2 (10 nodes, 15 links), RNP (27 nodes, 33 links) and GÉANT (40 nodes, 61 links) – Controllers capacity: 1800 kilorequests/s – Forwarding devices requests: 200 kilorequests/s – Percentage of controller backup resources: 30%

Comparison method

– Resilient placement strategy Zhang et al., denoted by MCC

[CUNHA et al., 2009; KNIGHT et al., 2011; TOOTOONCHIAN et al., 2012; ZHANG et al., 2011]

slide-25
SLIDE 25

Lucas F. Müller Survivor: Controller Placement Survivability – 25

Methodology

Four metrics

– Resilience

  • Resilience equation used by Zhang et al., 2011
  • Cardinal of edge-connectivity

– Overload

  • Number of overloaded controllers
  • Load distribution for each of the controller instances

[CUNHA et al., 2009; KNIGHT et al., 2011; TOOTOONCHIAN et al., 2012; ZHANG et al., 2011]

slide-26
SLIDE 26

Lucas F. Müller Survivor: Controller Placement Survivability – 26

Results: resilience

Probability of connectivity loss

(Resilience equation, Zhang et. al)

Survivor reduces the probability of connectivity loss.

0.1 0.2 0.3 0.4 0.5 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

Probability of connectivity loss Failure Probability

SVVR GEANT MCC RNP INTERNET2 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

Probability of connectivity loss Failure Probability

SVVR GEANT MCC RNP INTERNET2

(a) 1% a 10% (b) 0% a 100%

gain gain

slide-27
SLIDE 27

Lucas F. Müller Survivor: Controller Placement Survivability – 27

Results: resilience

Effect of exploring path diversity

(Cardinal of edge-connectivity)

Path diversity increases the network survivability, and it requires explicit consideration to be fully explored.

0.2 0.4 0.6 0.8 1 5 10 15 20

% of failure scenarios # of Disconnected elements

SVVR MCC

1 3 6

0.2 0.4 0.6 0.8 1 2 4 6 8 10

% of failure scenarios # of Disconnected elements

SVVR MCC+

1 3 6

CDFs of disconnected devices for all possible cases of 1, 3 and 6 link disruptions

slide-28
SLIDE 28

Lucas F. Müller Survivor: Controller Placement Survivability – 28

Results: overload

Number of overload scenarios

Network convergence after disruptions is highly sensible to predefined information in failover mechanisms.

5 10 15 20 SVVR MCC SVVR MCC SVVR MCC

# of overload scenarios

Normal Operation Failover strategy: Proximity Failover strategy: Residual Capacity

slide-29
SLIDE 29

Lucas F. Müller Survivor: Controller Placement Survivability – 29

Results: overload

Network state after convergence

(Load distribution)

Controller overload can be handled proactively by adding capacity-awareness and setting backup resources.

50 100 150 200 250 300 C1 C2 C3 C4 C5 C6 C7 C1 C2 C3 C4 C5 C6 C7

Load (%) SVVR MCC

50 100 150 200 250 300 C1 C2 C3 C4 C5 C6 C7 C1 C2 C3 C4 C5 C6 C7

Load (%) SVVR MCC

(b) Residual Capacity heuristic (a) Proximity heuristic

slide-30
SLIDE 30

Lucas F. Müller Survivor: Controller Placement Survivability – 30

Outline

  • Introduction: context and motivation
  • Proposed Approach: strategy and modeling
  • Results: resilience and overload
  • Conclusion
slide-31
SLIDE 31

Lucas F. Müller Survivor: Controller Placement Survivability – 31

Final Remarks

Contributions

– Significant reduction on connectivity loss – More realistic controller placement strategy – Smarter recovery mechanisms – Optimization model in order to generate optimal results

Ongoing work

– Studying meta-heuristics – Extend evaluation

slide-32
SLIDE 32

Lucas F. Müller Survivor: Controller Placement Survivability – 32

Contact Lucas Fernando Müller http://inf.ufrgs.br/~lfmuller lfmuller@inf.ufrgs.br

Survivor: an Enhanced Controller Placement Strategy for Improving SDN Survivability

Thank You!