Service Chain Placement in SDNs Gilad Kutiel 1 Dror Rawitz 2 1 - PowerPoint PPT Presentation

Service Chain Placement in SDNs Gilad Kutiel 1 Dror Rawitz 2 1 Technion 2 Bar Ilan University ALGOCLOUD 2017 1/15

Service Chain (Customer’s Perspective) � � � ◮ An ordered set of Virtual Functions � � � � 2/15

Service Chain (Customer’s Perspective) � � 100 mb/s � ◮ An ordered set of Virtual Functions ◮ Bandwidth requirement � � 60 mb/s � � 2/15

Service Chain (Customer’s Perspective) � � 100 mb/s � ◮ An ordered set of Virtual Functions ◮ Bandwidth requirement 90 ms � ◮ Latency constraint � 60 mb/s � � 2/15

Service Chain (Customer’s Perspective) � � 100 mb/s � � ◮ An ordered set of Virtual Functions ◮ Bandwidth requirement 90 ms � ◮ Latency constraint ◮ Several alternatives � � 60 mb/s � � � 2/15

Problem ( ISP’s Perspective) SDN � � � � � � � � � � � � � � � � � � � Service Chain 3/15

Problem ( ISP’s Perspective) SDN � � � � � � � � � � � � � � � � � � � Service Chain 3/15

Problem ( ISP’s Perspective) SDN � � ≥ 5 CPU � � � � � ≥ 8 CPU � � ≥ 12 CPU � � 3 CPU 5 CPU 5 CPU � � 2 CPU � � � � � 4 CPU 6 CPU � Service Chain 3/15

Problem ( ISP’s Perspective) SDN � � ≥ 80 mb/s � � � ≥ 80 mb/s � � ≥ 40 mb/s � � ≥ 40 mb/s � 80 mb/s � � � � � � � � 40 mb/s � Service Chain 3/15

Problem ( ISP’s Perspective) ≤ 90 ms SDN � � � � � � � � � � � � � � � � � � � Service Chain 90 ms 3/15

Problem ( ISP’s Perspective) ≤ 90 ms SDN � � 9$ � � � � � 7$ � � 4$ � � � � � � � � � � Service Chain 90 ms 3/15

Problem cont. Minimize Placement Cost s.t. ◮ CPU ◮ Bandwidth ◮ Latency 4/15

Related Work ◮ System 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem ◮ T. Lukovszki and S. Schmid 2015 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem ◮ T. Lukovszki and S. Schmid 2015 ◮ Even et al. 2016 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem ◮ T. Lukovszki and S. Schmid 2015 ◮ Even et al. 2016 ◮ Different Model / Objective 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem ◮ T. Lukovszki and S. Schmid 2015 ◮ Even et al. 2016 ◮ Different Model / Objective ◮ Even et al. 2016 5/15

Related Work ◮ System ◮ A. Gember-Jacobson et al. 2014 ◮ R. Hartert et al. 2015 ◮ Online Service Chain Embedding Problem ◮ T. Lukovszki and S. Schmid 2015 ◮ Even et al. 2016 ◮ Different Model / Objective ◮ Even et al. 2016 ◮ Cohen et al. 2015 5/15

Our Results ◮ NP-hardness in many cases 1 6/15

Our Results ◮ NP-hardness in many cases SDN Costs Our Result 1 6/15

Our Results ◮ NP-hardness in many cases SDN Costs Our Result DAG Integral, Polynomial Optimal 1 6/15

Our Results ◮ NP-hardness in many cases SDN Costs Our Result DAG Integral, Polynomial Optimal DAG Any FPTAS - (1 + ε )-apx 1 6/15

Our Results ◮ NP-hardness in many cases SDN Costs Our Result DAG Integral, Polynomial Optimal DAG Any FPTAS - (1 + ε )-apx Any Any FPTAS - (1 + ε )-apx 1 6/15

Our Results ◮ NP-hardness in many cases SDN Costs Our Result DAG Integral, Polynomial Optimal DAG Any FPTAS - (1 + ε )-apx FPTAS - (1 + ε )-apx 1 Any Any 1 running time depends on SDN’s typology 6/15

Hardness (Feasibility) 7/15

Hardness (Feasibility) � � 1 CPU 1 CPU � 1 CPU � 1 CPU � 1 CPU � 1 CPU � � � � � � 1 CPU 1 CPU 1 CPU 1 CPU 1 CPU 1 CPU Hamiltonian path ⇐ ⇒ Feasible Placement 7/15

Hardness (SDN with a single node) 8/15

Hardness (SDN with a single node) { 2 , 3 , 6 , 7 , 4 , 8 } 15 CPU � � � 2$ / 0 CPU 3$ / 0 CPU 6$ / 0 CPU 7$ / 0 CPU 4$ / 0 CPU 8$ / 0 CPU � � � � � � � � � � � � � � 0$ / 2 CPU 0$ / 3 CPU 0$ / 6 CPU 0$ / 7 CPU 0$ / 4 CPU 0$ / 8 CPU Partition ⇐ ⇒ Placement that costs 15$ 8/15

Hardness (SDN with a single node) { 2 , 3 , 6 , 7 , 4 , 8 } 15 CPU � � � 2$ / 0 CPU 3$ / 0 CPU 6$ / 0 CPU 7$ / 0 CPU 4$ / 0 CPU 8$ / 0 CPU � � � � � � � � � � � � � � 0$ / 2 CPU 0$ / 3 CPU 0$ / 6 CPU 0$ / 7 CPU 0$ / 4 CPU 0$ / 8 CPU Partition ⇐ ⇒ Placement that costs 15$ * Not hard for integral, polynomial values 8/15

Hardness (simple paths) 9/15

Hardness (simple paths) � � · · · · · · � � � � � � � · · · � � � � � · · · 9/15

Hardness (simple paths) � � · · · · · · � � � � � � � · · · � � � � � · · · See paper... 9/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 � � 1$ 3 CPU 6 0 � 10 CPU 5 0 � � 4 0 3 CPU 2 CPU 1$ 2$ � 3 0 2 0 � � 1$ 3 CPU 2$ 1 CPU 1 0 0 0 � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 � � 1$ 3 CPU 6 0 � 10 CPU 5 0 � � 4 0 3 CPU 2 CPU 1$ 2$ � 3 0 2 0 � � 1$ 3 CPU 2$ 1 CPU 1 0 0 0 � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 � � 1$ 3 CPU 6 0 � 10 CPU 5 0 � � 4 0 3 CPU 2 CPU 1$ 2$ � 3 0 2 0 � � 1$ 3 CPU 2$ 1 CPU 1 0 0 0 � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 � � 1$ 3 CPU 6 0 3 � 10 CPU 5 0 3 � � 4 0 3 3 CPU 2 CPU 1$ 2$ � 3 0 3 2 0 3 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 0 0 - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 � � 1$ 3 CPU 6 0 3 � 10 CPU 5 0 3 � � 4 0 3 3 CPU 2 CPU 1$ 2$ � 3 0 3 2 0 3 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 0 0 - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 � � 1$ 3 CPU 6 0 3 � 10 CPU 5 0 3 � � 4 0 3 3 CPU 2 CPU 1$ 2$ � 3 0 3 2 0 3 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 0 0 - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 � � 1$ 3 CPU 6 0 3 6 � 10 CPU 5 0 3 6 � � 4 0 3 6 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 2 0 3 6 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - 0 0 - - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 � � 1$ 3 CPU 6 0 3 6 � 10 CPU 5 0 3 6 � � 4 0 3 6 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 2 0 3 6 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - 0 0 - - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 5 � � 1$ 3 CPU 6 0 3 6 � 10 CPU 5 0 3 6 � � 4 0 3 6 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 2 0 3 6 � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - 0 0 - - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 5 � � 1$ 3 CPU 6 0 3 6 5 � 10 CPU 5 0 3 6 5 � � 4 0 3 6 5 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 5 2 0 3 6 - � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - - 0 0 - - - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 5 � � 1$ 3 CPU 6 0 3 6 5 � 10 CPU 5 0 3 6 5 � � 4 0 3 6 5 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 5 2 0 3 6 - � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - - 0 0 - - - � 1$ 3 CPU 10/15

Integral Polynomial Costs, Single Node Min CPU per Cost Budget. � � � � � � � � $ 7 0 3 6 5 9 � � 1$ 3 CPU 6 0 3 6 5 9 � 10 CPU 5 0 3 6 5 9 � � 4 0 3 6 5 9 3 CPU 2 CPU 1$ 2$ � 3 0 3 6 5 9 2 0 3 6 - - � � 1$ 3 CPU 2$ 1 CPU 1 0 3 - - - 0 0 - - - - � 1$ 3 CPU 10/15