Introduction to WAN Protocols

(1)

1

WMS-101 3154_06_2001_X

Introduction to WAN

Protocols

Session: WMS

Session: WMS-

-101

101

(2)

3 WMS-101

Technology Assumptions

• Basic Understanding of the OSI Reference Model

• Basic understanding of routing and switching.

• Basic Understanding of Networking Terms &

Acronyms

4 WMS-101

Definition of a Wide Area Network

A WAN is a network that covers a

broad geographic

area

and often uses transmission facilities provided

by

common carriers

. WAN technologies function at

the

lower three layers

of the OSI reference model:

• Physical Layer (L1)

• Data Link Layer (L2)

• Network Layer (L3)

(3)

5 WMS-101

Layer 2 Encapsulation

L2 Encapsulation

CPE A Router / WAN CPE B

Switch X

Time Division Multiplexing (TDM)

HDLC

HDLC PPPPPP FR / Frame Switching

FR / Frame Switching ATM / Cell Switching ATM / Cell Switching Application Presentation Session Transport Network Link Physical Link Physical Network Application Presentation Session Transport Network Link Physical

OSI Reference

Module

Layer 2 Frames:

• Transport for L3 across L1 • Error Detection & Possible Correction

• Establish peering across links • Different Characteristics

Why Understanding Protocols

Matters?

• Availability

• Scalability

• Efficiency

• Security

• Life Cycle

• Cost

Multiservice

Data Only

(4)

7 WMS-101

Making The Grade

M ult i-se rv_ic e M ult i-se rv_ic e D ata O_n ly D ata O_n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve ra_ll O ve ra_ll Protocol

Protocol _• _{Look at the technology in}

terms of individual requirements.

• Think of long term requirements ( 18mos - 3 years ) • Consider if protocol overhead or protocol delay is of more importance? 8 WMS-101

Do You Remember?

• What are the important characteristics to consider in

evaluating WAN protocols?

• What are the 3 HDLC Frame Formats?

• What are two applications of the Multilink Protocol in

PPP?

• What equivalent FRF specs exist in Frame Relay?

• In Frame Relay, what is the purpose of the FECN and

BECN bits in the Frame Header?

• What is one of the primary functions of the ATM

Adaption Layer?

(5)

9 WMS-101

Agenda

• Introduction

• Time Division Multiplexing (TDM)

• High Level Datalink Control (HDLC)

• Point to Point Protocol (PPP)

• Frame Relay (FR)

• Asynchronous Transfer Mode (ATM)

• Summary

Time Division Multiplexing (TDM)

• Timeslots are always present regardless if data is being sent.

D D D D D D D D E E E E D E E D E E MUX MUX

• Protocol Independent (HDLC, PPP, etc.)

• Bandwidth is statically allocated to the applications

• T1 (1.54Mbps) = 24 DSO’s or Channels of 64kbps each TS1

8 bits per timeslot

Framing (1 bit) 193 bits per frame (24*8 + 1) — 125 µµµµsec

TS3

(6)

11 WMS-101

TDM

TDM -

-

Application / ISDN

• Call Oriented Setup (Q.931)

• Fixed Bandwidth (No More / No Less)

• LAP D Frame Format (similar to HDLC)

64Kbps

ISDN BRIs Switch Switch

PSTN / ISDN To Corporate Network 64Kbps T1 = 24 DS0’s 12 WMS-101

Making The Grade

M ult i-se_rv ic_e M ult i-se rv_ic e D ata O n ly D ata_O n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve_ra ll O ve ra_ll Protocol Protocol

• Definite Support for Multiservice Applications.

• Predictable Delay

• Bandwidth likely to be under-utilized.

• Secure, L1 End-to-End

• Will be around for a while, but likely usurped by converged networks.

• Costs can be prohibitive in a tariffed environment.

A

C

D

C

D+

C

D

C

A

(7)

13 WMS-101

Agenda

• Introduction

• Time Division Multiplexing (TDM)

• High Level Datalink Control (HDLC)

• Point to Point Protocol (PPP)

• Frame Relay (FR)

• Asynchronous Transfer Mode (ATM)

• IP-VPNs

• Summary

HDLC

• HDLC supports 16 or 32 bit Checksums

• HDLC supports 3 modes; NRM, ARM,

and ABM

• HDLC LAP B is the WAN relevant

application

• HDLC is sequenced and can perform

Flow and Error control

(8)

15 WMS-101

L3 Datagram

HDLC

HDLC -

-

Frame Format

• 3 Frame Types: Information, Supervisory, & Unnumbered

• Point-to-Point configuration typically employed

• Cisco HDLC (proprietary) Point-to-Point Configuration

(Data) L3 Datagram

Control

Flag Address FCS Flag 1 1 or 2 1 or 2 Variable 2 1 I-Frames N(R) P N(S) 0 0x0F 0x00 0x0800 Cisco Frame OR 16 WMS-101

HDLC

-

Application

• Point-to-Point Applications (Leased Line)

• L2 QoS Doesn’t Matter / Data Throughput Matters

• No Multiservice L2 Intelligence / L3 Queuing can partially assist

• Under-utilized links makes Multiservice possible on High Speed links (DS3+), but unpredictable.

(9)

17 WMS-101

Making The Grade

M ult i-se rv_ic e M ult i-se rv_ic e D ata O_n ly D ata O_n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve ra_ll O ve ra_ll Protocol Protocol

• Data Only = Excellent

• Currently supported up to DS3 links, with rate limiting for sub.

• Light Overhead, ideal for applications where maximum throughput matters.

A

B

-

-B

B

D

A

B-

B

--

-

-A

A

C

F

D

Agenda

• Introduction

• Time Division Multiplexing (TDM)

• High Level Datalink Control (HDLC)

• Point to Point Protocol (PPP)

• Frame Relay (FR)

• Asynchronous Transfer Mode (ATM)

(10)

19 WMS-101

PPP

• Point-to-Point Protocol. Used in Dial, xDSL, ISDN,

Serial applications

• PPP can Multiplex multiple Network Protocols over a

single link (Protocol Agnostic)

• Options for IP address assignment and management

• Link Configuration, Quality, and Error Detection

• Can negotiate additional options for Authentication,

Compression, Multilink Support, etc.

• PPP uses an HDLC Frame for Encapsulation

20 WMS-101

PPP

PPP -

-

Frame Format

L3 Datagram 0x03 Flag 0xFF FCS Flag 1 1 1 2 0 - 1500 2 1 0x0800 PPP doesn’t assign individual station address

therefore using the broadcast address

Indicates transmission of user data in an

non-sequenced frame (connectionless)

Indicates the NLPID of the L3 Datagram in the payload of the frame

Maximum Transmission Unit (minus overhead)

CRC Error Checking Protocol ID’s Novell Appletalk 0x809B 0x8137 NetBIOS 0x00F0 0x00BC 0x0000 Banyan More..

(11)

21 WMS-101

PPP

PPP -

-

Operation

LCP: • LCP Listen • Option Negotiation • Link Quality is determined (Optional) • Network Layer Configuration Begins (IPCP, IPXCP, ATCP) • Link Establishment

(LCP Open) • LCP Termination

Se2/0:7 PPP: Phase is ESTABLISHING, Passive Open [0 sess, 0 load] Se2/0:7 LCP: State is Listen

Se2/0:7 LCP: I CONFREQ [Listen] id 230 len 27 Se2/0:7 LCP: AuthProto CHAP (0x0305C22305)

Se2/0:7 LCP: MagicNumber 0x4CDA0A5B (0x05064CDA0A5B) Se2/0:7 LCP: MRRU 1524 (0x110405F4)

Se2/0:7 LCP: EndpointDisc 1 1720a (0x1308013137323061)

Se2/0:7 LCP: I CONFACK [ACKsent] id 76 len 30 Se2/0:7 LCP: AuthProto CHAP (0x0305C22305)

Se2/0:7 LCP: MagicNumber 0xCC96D7E6 (0x0506CC96D7E6) Se2/0:7 LCP: MRRU 1524 (0x110405F4)

Se2/0:7 LCP: EndpointDisc 1 3640_PE1 (0x130B01333634305F504531) Se2/0:7 LCP: State is Open

Se2/0:7 LCP: O CONFREQ [Listen] id 76 len 30 Se2/0:7 LCP: AuthProto CHAP (0x0305C22305)

Se2/0:7 LCP: MagicNumber 0xCC96D7E6 (0x0506CC96D7E6) Se2/0:7 LCP: MRRU 1524 (0x110405F4)

Se2/0:7 LCP: EndpointDisc 1 3640_PE1 (0x130B01333634305F504531) Se2/0:7 LCP: O CONFACK [Listen] id 230 len 27

Se2/0:7 LCP: AuthProto CHAP (0x0305C22305) Se2/0:7 LCP: MagicNumber (0x05064CDA0A5B) Se2/0:7 LCP: MRRU 1524 (0x110405F4)

Se2/0:7 LCP: EndpointDisc 1 1720a (0x1308013137323061)

PPP

PPP -

-

Authentication (CHAP)

CHAP Characteristics:

• 3-Way Handshake on link establishment.

• Authenticator sends a “Challenge”

• Peer responds with a value based on a one-way hash

• Authenticator validates against its own calculation.

Se2/0:7 CHAP: I CHALLENGE id 69 len 26 from "1720a" Se2/0:7 CHAP: Waiting for peer to authenticate first

Both Peers Challenging (Debug):

1720a

3640a

Se2/0:7 CHAP: I RESPONSE id 76 len 26 from "1720a" Se2/0:7 PPP: Phase is FORWARDING [0 sess, 0 load] Se2/0:7 PPP: Phase is AUTHENTICATING [0 sess, 0 load] Se2/0:7 PPP: Phase is AUTHENTICATING, by both [0 sess, 0 load] Se2/0:7 CHAP: O CHALLENGE id 76 len 29 from "3640a"

Se2/0:7 CHAP: O SUCCESS id 76 len 4 Se2/0:7 CHAP: Processing saved Challenge, id 69 Se2/0:7 CHAP: O RESPONSE id 69 len 29 from "3640a" Se2/0:7 CHAP: I SUCCESS id 69 len 4

(12)

23 WMS-101

PPP

PPP -

-

NCP Negotiation

NCP Characteristics:

• Responsible for configuring, enabling and disabling the L3 protocol.

• Uses L2 protocol field 0x8021 to identify the payload as IPCP

• Address Assignment (DHCP)

• NetBios Name Servers

• Domain Name System

Holder

Both Peers Challenging (Debug):

1720a

3640a e:

Need debug output, but lab is tore down until June 4th. e:

Need debug output, but lab is tore down until June 4th.

24 WMS-101

PPP

PPP -

-

Multilink

• LCP Negotiated Option

• Member Links Identified through Endpoint Discriminator and / or Authenticated name.

• Bundles Multiple Physical Links into a logical bundle

• Bandwidth on Demand

(13)

25 WMS-101

PPP

PPP -

-

Fragmentation &

Interleaving

• MP Fragmentation Breaks up Large Data Packets in smaller sequenced fragments.

• Fragment-Delay is used to stipulate the maximum time a fragment can be on an individual link

• MP creates opportunities for non-MP encapsulated traffic (I.e, RTP) used in Voice applications to be interleaved.

• MP fragmentation and interleaving ideal in low speed (< 1.2Mbps) where delay is priority over throughput.

Making The Grade

M ult i-se_rv ic_e M ult i-se rv_ic e D ata O n ly D ata_O n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve_ra ll O ve ra_ll Protocol

Protocol • Primarily used in Data

applications, however, can be used from Multiservice

• Mature Protocol with new life in Broadband

Aggregation applications

• HDLC style header is efficient for Data, MP is efficient for Multiservice | BW Aggregation.

A

B-

B

A

B

-A

A

--

B

B+

--

-

B-B-

B

(14)

27 WMS-101

Agenda

• Introduction

• Time Division Multiplexing (TDM)

• High Level Datalink Control (HDLC)

• Point to Point Protocol (PPP)

• Frame Relay (FR)

• Asynchronous Transfer Mode (ATM)

• Summary

28 WMS-101

Frame Relay

Frame Relay -

-

Overview

What is the purpose / advantage of a ‘Virtual Circuit’?

Chicago New York Miami Dallas San Francisco

**N * (N-1) / 2 = Full Mesh**

5 Sites = 10 LL

10 Sites = 45 LL

(15)

29 WMS-101

Frame Relay

• Packet Switched (Compared to Circuit Switched)

• Statistical Multiplexing Alleviates Wasted ‘silence’

• Uses a Virtual Circuit (VC) or Path through the network

• BW is not Allocated Until Needed

• Buffering and Congestion Control mechanisms

• Relies on Upper Layer Protocols (ex. TCP) for error recovery

• Frame Relay supported up to 45Mbps

SP Network

www.corporate.com HQ

Branch

Frame Relay

Frame Relay -

-

Frame Format

• DLCI - 10 Bit field (1024 Possible connections), Locally Significant

L3 Datagram (Data) Flag FCS Flag 1 Variable ( 0 ~ 4096) 2 1 Header DLCI C/R 6 1 EA FEC_N 1 1 B_E C_{N DE} 1 1 EA 2 Bits 1 4 DLCI • C / R - Undefined Field

• EA - Extended Address ( 1 = End, 0 = More DLCI in 2nd Octet)

• FECN - Forward Explicit Congestion Notification ( --> Direction)

• BECN - Backward Explicit Congestion Notification ( <-- Direction)

• DE - Discard Eligibility: Set by end node allows frames to be Bytes

(16)

31 WMS-101

Frame Relay

Frame Relay -

-

LMI

• LMI - Local Management Interface – VC Discovery (DLCI)

– Multicasting – Global Addressing

• LMI is used to check the ‘Status’ of PVCs on the network

• LMI Uses reserved DLCI ( 0 = ITU, ANSI or 1023 = Cisco)

DCE DTE Status _Enquir y Status

Which DLCI’s

are active?

Frame DLCI 19, 23, 58 = ActiveDLCI 21, 29, 5 = Inactive

Switch Enquiry Types: • Short • Long • Asynchronous 32 WMS-101

Frame Relay

Frame Relay -

-

UNI / NNI

• The Service Provider’s cloud could be non-FR (I.e. ATM, etc.)

• Inverse ARP allows Network Layer address discovery (RFC 1293)

• Static Mapping required without use of iARP (not manageable)

• DLCI’s are Locally significant. DLCI swapping is job of FR the Switch.

• The SP network will set FECN & BECN bits based on Congestion

• The SP will set DE bits based on Service Contracts.

Service Provider Cloud Site A Site C Site B DLCI 100 DLCI 120 DLCI 60 DLCI 80 NNI UNI UNI UNI Frame Switch Frame Switch

(17)

33 WMS-101

Frame Relay

Frame Relay -

-

FRF.12

• Fragment Large frames into a sequence of shorter frames

• Control Delay ‘critical’ for Multiservice applications (Voice, etc.)

• Fragmentation occurs on a per-VC basis

• 2 Byte Sequence Header keeps packets ordered (10 bits seq.)

• Large Frames hog time on wire, create delay problems

PHY I/F V V V V Data Data PHY I/F DLCI 120 DLCI 100 DTE - DTE Fragmentation V V V V Data Data DLCI 54 DLCI 147 V D D V D V D V D

Frame Relay

Frame Relay -

-

FRF.16

• FRF16 = Multilink Frame Relay

• Same encapsulation as FRF12 - UNI / NNI Fragmentation

• Increase Bandwidth where there are service offering gaps (T1 x N)

• Eliminate single points of failure with Physical interfaces.

• Inverse MUX’ing several Physical Interfaces into 1 Logical Interface

PHY I/F PHY I/F PHY I/F PHY I/F MFR MFR Data 2 Data 1 Data 1 Data 2 Data 1 Data 2

(18)

35 WMS-101

Frame Relay

Frame Relay -

-

Application

Chicago New York Head Quarters Dallas San Francisco DLCI 31 DLCI 32 DLCI 33 DLCI 34

• Sales / Remote

Offices

• Reduces Interfaces

• Simplify

Configuration

• Partial Mesh or

Hub and Spoke

design

• Reduce LL costs

DLCI 1 DLCI 2 DLCI 3 DLCI 4 36 WMS-101

Frame Relay

Frame Relay -

-

Characterization

•

• Dynamic Allocation of Bandwidth Dynamic Allocation of Bandwidth •

•Bandwidth is not wasted. The hungry mouth gets it.Bandwidth is not wasted. The hungry mouth gets it.

•

•Statistical Multiplexing allows idle VC’s to share bandwidth wiStatistical Multiplexing allows idle VC’s to share bandwidth with active VC’sth active VC’s

•

• Can Be Used for Multiservice ApplicationsCan Be Used for Multiservice Applications

Frame Switches Are Used for Multiservice Applications (DVV) (Les

Frame Switches Are Used for Multiservice Applications (DVV) (Less over s over

Subscription and Reasonable Speed Links)

•

• Technology is still being enhanced (FRF.12, FRF.16, etc.)Technology is still being enhanced (FRF.12, FRF.16, etc.)

•

• Bandwidth is expandable (FRF.16)Bandwidth is expandable (FRF.16)

•

• Unable to Guarantee Performance (in FIFO Mode)Unable to Guarantee Performance (in FIFO Mode)

Frame Switches ‘Typically’ Operate in FIFO (First in

Frame Switches ‘Typically’ Operate in FIFO (First in--First out) Mode, so One First out) Mode, so One Application Can Impact the Performance of Others

Application Can Impact the Performance of Others

•

• Medium Delay and Variability in DelayMedium Delay and Variability in Delay

Each Switch Has to Receive an Entire Frame before Forwarding It

Each Switch Has to Receive an Entire Frame before Forwarding It to the Next to the Next

Switch; Therefore Transit Delay Increases with Number of Switche

Switch; Therefore Transit Delay Increases with Number of Switches in the Paths in the Path The FIFO Mode of Each Switch Causes a Variability at Each Switch

The FIFO Mode of Each Switch Causes a Variability at Each Switch

Cons

Pros

(19)

37 WMS-101

Frame Relay: Report Card

M ult i-se rv_ic e M ult i-se rv_ic e D ata O_n ly D ata O_n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve ra_ll O ve ra_ll Protocol Protocol

• Fits into a Multiservice Application.

• Speeds up to DS3 and MFR scales (NxT1).

• Light Protocol Overhead (2 Bytes) and LFI make it efficient for Data and Multiservice.

A

B

B+

B

-

B

A

B

B+

-L2

L2

B+

-

B+

B

B-B-

B+

B

Agenda

• Introduction

• Time Division Multiplexing (TDM)

• High Level Datalink Control (HDLC)

• Point to Point Protocol (PPP)

• Frame Relay (FR)

• Asynchronous Transfer Mode (ATM)

(20)

39 WMS-101

ATM

ATM -

-

Overview

• Connection Oriented transport (VC’s pre-established)

known as ‘Cell Switching’

• Hybrid of Circuit Switching and Packet Switching

• Fixed Cell size 5byte Header + 48byte Payload

reduces latency typical to large data packets

• ATM Supports Multiple Qualities of Service

• Virtual Path + Virtual Channel = Virtual Circuit

• ATM supports Permanent VC’s and Switched VC’s

• ATM speeds up to OC-48 (2.5Gbps)

40 WMS-101

ATM

ATM -

-

Functional Layers

Physical Layer 1 ATM Layer AAL Physical Layer Data Link Layer Network Layer OSI RM B-ISDN RM

• Segmentation & Reassembly

• Payload Error Control

• End-to-End Timing

• VPI / VCI Switching

• Cell MUX / DEMUX

• Flow Control / HEC

• QoS Support

• Bitstream Conversion

(21)

41 WMS-101

ATM

ATM -

-

Cell Format

GFC - Generic Flow Control VPI - Virtual Path Identifier VCI - Virtual Channel Identifier PT - Payload Type

CLP - Cell Loss Priority HEC - Header Error Check

H P H P H P H P H P

Header Payload

Transmission Path

GFC VPI VCI PT CLP HEC

4 8 16 3 1 8

ATM

ATM -

-

Operation

IP Datagram LLC

Segmentation

ATM Adaption Layer

ATM Layer

PHY Layer IP Data 48 48 48

VPI / VCI Assignment

5 5 5 5 5 5 VC MUXing 5 5 5 Serialization 101100111010110011000100111101100

(22)

43 WMS-101

ATM

ATM -

-

Traffic Definitions

• CBR - Constant Bit Rate, Connection Oriented w /

end-to-end timing required, utilizes AAL1 (Leased

Line Emulation)

• ABR - Available Bit Rate

• UBR - Unspecified Bit Rate, connectionless packet

data, ‘best-effort’ transport. No guarantees to loss,

delay, or bandwidth available, utilizes AAL5

• Others, VBR-NRT, VBR-RT, etc.

44 WMS-101

ATM

ATM -

-

Application

New York Head Quarters San Francisco

• Enterprise WAN

Core

• Define Multiple

Traffic Contracts

• Predictable Delays

for Multiservice

Applications

• No under-utilized

bandwidth (like

TDM)

• Scale VC’s by

application.

Data - AAL5 UBR Voice - AAL1 CBR

(23)

45 WMS-101

Public ATM Switch

ATM I

ATM I-

-LMI

LMI

• Integrated local management interface-ilmi

• Use SNMP across UNI and NNI for ILMI MIB

• Uses AAL 5 encapsulation

• Used for ATM end system address (AESA) formerly NSAP addressing for svc’s

• Automatic recognition of UNI or NNI interface protocol

Site A Site C Site B UNI ATM Switch ATM Switch End-System Private / Public

Switch _{Private / Public}

Switch End-System IME IME IME IME IME

Making The Grade

M ult i-se_rv ic_e M ult i-se rv_ic e D ata O n ly D ata_O n ly Scalability Scalability Efficiency Efficiency Security Security Cost Cost Life Cycle Life Cycle Availability Availability O ve_ra ll O ve ra_ll Protocol Protocol

• ATM is great for

multiservice applications, data-only pays a cell tax

• Bandwidth is scalable up to 2.5Gbps

• Delay is predictable and bandwidth use is efficient, more applications coming

B

A

B-B-

B+

--

-

-C+

C+

D

-

-B

B

(24)

47 WMS-101

ATM

ATM -

-

Characterization

•

•Dynamic Allocation of BandwidthDynamic Allocation of Bandwidth

Available Bandwidth Is Allocated Dynamically to Any Application

Available Bandwidth Is Allocated Dynamically to Any Application that Needs Itthat Needs It One Application Can Use Bandwidth Allocated to the other if that

One Application Can Use Bandwidth Allocated to the other if thatTraffic Is Not PresentTraffic Is Not Present •

•Guaranteed performanceGuaranteed performance

Cell Switches with Efficient Traffic and Bandwidth Management Sc

Cell Switches with Efficient Traffic and Bandwidth Management Schemes Can Ensure thathemes Can Ensure that Each Application Receives Guaranteed Performance (TM,

Each Application Receives Guaranteed Performance (TM,QoSQoSQueuing, CAC, PNNI/UNI Etc.)Queuing, CAC, PNNI/UNI Etc.) •

•Low Delay (Controlled and Bounded) and Low Variability in DelayLow Delay (Controlled and Bounded) and Low Variability in Delay

Using Fixed Length Cells Ensures that Network Transit Delay and

Using Fixed Length Cells Ensures that Network Transit Delay and Variability in DelayVariability in Delay Is Minimized

Is Minimized Switches Use

Switches UseQoSQoS--Based Queuing and Scheduling Such as CBR, VBR, ABR Based Queuing and Scheduling Such as CBR, VBR, ABR •

•TypicallyTypicallyMultiserviceMultiservice

As a Result of Low Delay, Low Variability in Delay and the Abili

As a Result of Low Delay, Low Variability in Delay and the Ability to Guarantee Performance, ty to Guarantee Performance, Cell Switches Are Ideally Suited to Support Multiple Services Co

Cell Switches Are Ideally Suited to Support Multiple Services Concurrentlyncurrently

•

• Overhead Overhead

However the Bandwidth Efficiency and Ability to Provide Low Dela

However the Bandwidth Efficiency and Ability to Provide Low Delay and y and Low Variability in Delay in Cell Switching Easily Overcomes the

Low Variability in Delay in Cell Switching Easily Overcomes the Small Small

Incremental Overhead Incremental Overhead

Cons

Pros

48 WMS-101

All Together Now

ATM Frame-Relay HDLC PPPoX New York Dallas San Francisco OC-3 Internet DS-3 Chicago Boston Austin Ft. Worth 384K _256K 256K 128K VPDN