stacked-vlan-based modeling of hybrid isp traffic control

Stacked-VLAN-Based Modeling ofHybrid ISP Traffic Control Schemesand Service Plans Exploiting Excess

Bandwidth in Shared AccessNetworks

Dr Kyeong Soo (Joseph) Kim

Department of Electrical and Electronic EngineeringXi’an Jiaotong-Liverpool University

15-16 September 2016

1 / 21

Outline

Introduction

Review of Hybrid ISP Traffic Control for Shared Access

Modeling of Hybrid ISP Traffic Control Schemes andService Plans based on Stacked VLANs

Summary

2 / 21

Next . . .

Introduction

Review of Hybrid ISP Traffic Control for Shared Access

Modeling of Hybrid ISP Traffic Control Schemes andService Plans based on Stacked VLANs

Summary

3 / 21

Current ISP Traffic Control Architecture

…

Egress

Classifier

TBF*

TBF

Scheduler

SNII/F with

(O)DNRelay Unit

MACPort

Filter

ISP

Per-Subscriber

Traffic

Control

MACPort

Filter

Access Switch

App.

Server

RB

RF

(=RD)

RD

RD

Subscriber

Unit 1

Subscriber

Unit N

…

Subscriber

1

RU

Access

Switch

Shared

(O)DN*

RU

RouterRB

SNI UNIAccess Network

Subscriber

N

…

* (O)DN: (Optical) Distribution Network

* SNI: Service Node Interface

* TBF: Token Bucket Filter

* UNI: User-Network Interface

4 / 21

Issues with Current ISP Traffic Control

The arrangement of traffic shapers and a schedulerprevents subscribers from sharing available bandwidth.

5 / 21

Toward Fully-Shared Access

To implement fully-shared access networks for betterresource utilization and higher energy efficiency, we havebeen studying the following:

ISP traffic control schemes enabling excessbandwidth allocation 12.Hybrid ISP traffic control architecture and serviceplans exploiting excess bandwidth 3.Implementation of simulation models for theproposed traffic control architecture and serviceplans (reported in this paper).

1K. S. Kim, IEEE Commun. Lett., vol. 18, no. 4, Apr. 2014.2K. S. Kim, Proc. IEEE Sarnoff Symposium 2015, Aug. 2015.3K. S. Kim, Proc. ICTC 2014, Busan, Korea, Oct. 2014.

6 / 21

Next . . .

Introduction

Review of Hybrid ISP Traffic Control for Shared Access

Modeling of Hybrid ISP Traffic Control Schemes andService Plans based on Stacked VLANs

Summary

7 / 21

Hybrid ISP Traffic Control Architecture I

…Egress

Classifier

Relay

Unit

Output

Port

TBF

TBF

Regular

Scheduler

…

TBM*

TBM

Scheduler

Allocating

Excess

Bandwidth

TBF

Virtual Subscriber

* TBM: Token Bucket Meter

8 / 21

Hybrid ISP Traffic Control Architecture II

Backward compatibility• A group of new service plan subscribers are treated as

one virtual subscriber under the existing flat-rateservice plan.

Better resource utilization• The new service plan subscribers can fully share the

bandwidth within the group.

Gradual introduction• Migration to a fully-shared access will be completed

when all the subscribers of the existing service planmove to the new one.

9 / 21

Next . . .

Introduction

Review of Hybrid ISP Traffic Control for Shared Access

Modeling of Hybrid ISP Traffic Control Schemes andService Plans based on Stacked VLANs

Summary

10 / 21

Modeling of Hybrid Traffic Control I

Major requirements:Backward compatibility with the existing VLANimplementations in INET-HNRL, including• EthernetFrameWithVLAN message format• MACRelayUnitNPWithVLAN and VLANTagger

modules

Expandability to stack more than two VLAN tags

11 / 21

Modeling of Hybrid Traffic Control II

Two possible approaches:Integrated approach• The whole scheduling is implemented as one

integrated scheduler like the hierarchical tokenbucket (HTB) scheduler 4.

Modular approach• Separate schedulers (e.g., a scheduler based on TBF

shaping and a DRR-based scheduler enabling excessbandwidth allocation 5) are combined into one.

4M. Devera. Linux HTB home page.5K. S. Kim, Proc. IEEE Sarnoff Symposium 2015, Aug. 2015.

12 / 21

Identification of Subscriber Frames IFor the existing traffic controlscheme, the whole framesfrom the subscribers of thehybrid scheme are identifiedand treated as a group (i.e.,one virtual subscriber).For the newexcess-bandwidth-allocatingtraffic control scheme, theframes from each subscriberare identified and treated as aseparate flow.This can be done by IEEE802.1Q stacked VLANs (alsocalled provider bridging andQ-in-Q). 13 / 21

Frame Formats for VLAN Stacking

Destination

Address

(DA)

Source

Address

(DA)

Ethertype Payload

Frame Check

Sequence

(FCS)

Destination

Address

(DA)

Source

Address

(DA)

802.1Q Header

(TPID=0x8100)

Destination

Address

(DA)

Source

Address

(DA)

802.1Q Header

(TPID=0x88A8)

802.1Q Header

(TPID=0x8100)

Tag

PushTag

Pop

Ethernet II Frame

Ethernet II Frame with Single VLAN TAG

Ethernet II Frame with Double VLAN TAGs

S-VLAN C-VLAN

Ethertype Payload

Frame Check

Sequence

(FCS)

Ethertype Payload

Frame Check

Sequence

(FCS)

14 / 21

Ethernet II Frame Message DefinitionsWithout VLAN stacking:packet EthernetIIFrameWithVLAN extends EthernetIIFrame{

uint16_t tpid = 0x8100; // tag protocol identifier (16 bits; set to 0x8100)uint8_t pcp; // priority code point for IEEE 802.1p class of service (3 bits; 0 (lowest) to 7 (highest))bool dei; // drop eligible indicator (1 bit)uint16_t vid; // VLAN identifier (12 bits; 0x000 and 0xFFF are reserved, which allows up to 4094 VLANs)

}

With VLAN stacking:cplusplus {{#include <stack>#include "VLAN.h" // define VLANTag structtypedef std::stack<VLANTag> VLANTagStack;}}

class noncobject VLANTagStack;

packet EthernetIIFrameWithVLAN extends EthernetIIFrame{

uint16_t tpid; // tag protocol identifier (16 bits; set to 0x8100 for C-TAG & 0x88A8 for S-TAG)uint8_t pcp; // priority code point for IEEE 802.1p class of service (3 bits; 0 (lowest) to 7 (highest))bool dei; // drop eligible indicator (1 bit)uint16_t vid; // VLAN identifier (12 bits; 0x000 and 0xFFF are reserved, which allows up to 4094 VLANs)

// optional; IEEE 802.1Q-in-Q stacked VLANs.VLANTagStack innerTags; // based on std::stack

}

15 / 21

Stacked-VLAN-Based Modeling of An AccessNetwork with Hybrid ISP Traffic Control

16 / 21

Modules with VLAN Support

17 / 21

Backpressure between SchedulersTwo schedulers are currently connected based onEthernet flow control, which halts the transmission ofthe whole frames.For better control, we need an extension (like IEEE802.1Qbb) to halt the transmission of non-conformantframes only.

18 / 21

Throughput Example

9 10 11 12 13 140

10

20

30

40

50

60

70

80

90

100

Th

rou

gh

pu

t [M

b/s

]

RR+TBF (Original)

9 10 11 12 13 14

Time [s]

1.11s

CSFQ+TBM

9 10 11 12 13 14

DRR+TBM (Proposed)

19 / 21

Next . . .

Introduction

Review of Hybrid ISP Traffic Control for Shared Access

Modeling of Hybrid ISP Traffic Control Schemes andService Plans based on Stacked VLANs

Summary

20 / 21

Summary

Discussed the issues in current practice of ISP trafficshaping and related flat-rate service plans in sharedaccess networks.Reviewed alternative service plans based on newhybrid ISP traffic control schemes exploiting excessbandwidth.Reported the current status of our modeling of thehybrid ISP traffic control schemes and service planswith OMNeT++/INET-HNRL based on stackedVLANs.

21 / 21