Chapter 9 ATM Networks

1
Chapter 9ATM Networks

• Why ATM?
• BISDN Reference Model
• ATM Layer
• ATM Adaptation Layer
• ATM Signaling
• PNNI Routing
• Classical IP over ATM

2
Chapter 9ATM Networks

• Why ATM?

3
The Integrated Services Vision

• Initially telephone network all-analog
• Transmission Switching
• Gradual transition to all-digital core
• 1960s transmission in backbone became digital
• 1970s switching became digital
• Subscriber loop from customer to network remained
  analog
• Integrated Services Vision
• Network should be digital end-to-end
• Network should support all services telephone,
  data, video
• Three attempts at achieving Integrated Services
  Network
• ISDN in 1980s
• ATM/BISDN in 1990s
• Internet in 2000s

4
Integrated Services Digital Network (ISDN)

• ISDN
• Integrated access
• to end-to-end digital communication services
• through a standard set of user-to-network
  interfaces
• Network consisted of separate networks for voice,
  data, signaling

Circuit- switched network
B64 kbps D16 kbps

• Basic rate interface (BRI) 2BD

Private channel- switched network
BRI
BRI
Packet- switched networks
PRI
PRI
Primary rate interface (PRI) 23BD
Signaling network
5
Broadband ISDN

• BISDN A single universal network that is
  flexible enough to provide all user services in a
  uniform manner
• ISDN not enough Needed 10s to 100s Mbps for LAN
  interconnect and for digital TV
• Synchronous Transfer Mode (connections at nx64
  kbps) was initial candidate for BISDN, but
• Asynchronous Transfer Mode (ATM) chosen
• Multiplexing switching framework
• connection-oriented virtual circuits
• fixed-length packets, cells, with short
  headers

6
Benefits of ATM

• Network infrastructure and management simplified
  by using a single transfer mode for the network
• Expected to cover LAN, MAN, and WAN
• Extensive bandwidth management capabilities
• SONET-like grooming capabilities, but at
  arbitrary bandwidth granularities
• ATM is not limited by speed or distance
  limitations
• 50-600 Mbps the sweet spot for ATM
• QoS attributes of ATM allow it to carry voice,
  data, and video thus making it suitable for an
  integrated services network.

7
ATM Anticipated Scope

• All information transferred by network that
  handles 53-byte cells
• Scalable in terms of speed
• Switched approach operates in LAN, MAN, or WAN

8
ATM Networking
9
AAL converts Info into Cells
10
Cell-Switching Virtual Circuit

• Connection setup establishes virtual circuit by
  setting pointers in tables in path across network
• All cells for a connection follow the same path
• Abbreviated header identifies connection
• Cells queue for transmission at ATM switches
  multiplexers
• Fixed and Variable bit rates possible, negotiated
  during call set-up
• Delay and loss performance negotiated prior to
  connection setup

11
ATM Switching
Switch carries out table translation and routing
ATM switches can be implemented using shared
memory, shared backplanes, or self-routing
multi-stage fabrics
12
Multiplexing in ATM Switches

• Packet traffic multiplexed onto input lines
• Demultiplexed at input port
• Forwarded to output port

13
ATM Support for Multiple QoS Levels
VCs with different TDs different QoS reqts

• Call Admission Control based on Traffic
  Descriptors QoS Reqts
• Cell streams policed at User Network Interface
• Cell Enqueueing Policy, Cell Transmission
  Scheduling, Flow Control
• Generalized Processor Sharing, Weighted Fair
  Queueing, etc.
• Multiplexing Gain
• Cell Multiplexing implies Delay, Jitter, Loss

14
Chapter 9ATM Networks

• BISDN Reference Model

15
BISDN Reference Model

• User Plane transfer of user information flow
  control error recovery
• Control Plane setting up, management, and
  release of connections
• Layer Management Plane management of layer
  entities OAM
• Plane Management management of all the planes

16
Planes Explained

• Three types of logical networks are involved in
  delivering communication services
• User Network transfers user information
• Control (Signaling) Network carries signaling
  messages to establish, maintain, terminate
  connections
• Management Network carries management
  information monitoring information, alarms and
  usage statistics
• A separate protocol stack, plane, is defined
  for each of these three networks

17
ATM Layered Architecture
18
ATM Layered Architecture

• ATM Adaptation Layer
• standard interface to higher layers
• adaptation functions
• end-to-end between end systems
• segmentation into cells and reassembly
• ATM Layer
• Transfer of Cells
• Cell-Header Generation/Extraction
• VPI/VCI Translation
• Cell multiplexing/demultiplexing
• Flow and congestion control
• Physical Layer
• Cell stream / bit stream conversion
• Digital transmission

19
ATM Interfaces
Private ATM network
UNI User-Network Interface NNI
Network-Network Interface B-ICI Broadband
Inter-carrier i/f
Private UNI
X
X
Private NNI
Public ATM network A
Public UNI
X
X
X
NNI
Public UNI
X
Public ATM network B
B-ICI
X
Public UNI
X
X
20
The ATM Physical Layer

• TC Sublayer
• Cell Delineation
• Header Error Checking
• Cell Rate Decoupling (Insertion of Idle Cells)
• Specific to PMD
• PMD Sublayer
• Line code
• Connectors
• Re-use of existing physical layer standards

21
Private UNI Physical Layers
UTP Unshielded twisted pair STP Shielded
twisted pair MMF Multimode fiber SMF
Single-mode pair
22
Public UNI Physical Layers
23
Chapter 9ATM Networks

• ATM Layer

24
The ATM Layer

• Concerned with sequenced transfer of cells across
  network connections
• ATM Connections
• Point-to-Point unidirectional or bidirectional
• Point-to-Multipoint unidirectional
• Permanent Virtual Connections (PVC) long-term
  connections to provision bandwidth between
  endpoints in an ATM network
• Switched Virtual Connections (SVC) shorter-term
  connections established in response to customer
  requests

25
ATM Virtual Connections

• Virtual Channel Connections virtual circuit
• Virtual Path Connections bundle of virtual
  connections
• ATM Header contains virtual connection
  information
• 8-bit Virtual Path Identifier
• 16-bit Virtual Channel Identifier

26
Why 53 Bytes?

• The effect of delay on packet voice influenced
  selection of cell size
• The packetization delay grows with the cell size
• _at_64kbps packetization delay cell size 125
  ?sec
• If delay is too long, echo cancellation equipment
  needs to be introduced
• Europe has short transmission lines and no echo
  cancellers so it proposed 32 byte payload
• U.S. has long transmission lines and echo
  cancellers in place, so it proposed 64 byte
  payload
• Compromise 48 byte payload

27
The ATM Cell

• Virtual Path Identifier
• 8-bits 256 VC bundles
• Virtual Channel Identifier
• 16 bits 65,536 VCs/VP
• Payload Type Indicator
• Bit 3 data vs. OAM cell
• Bit 2 Congestion indication in data cells
• Bit 1 Carried transparently end-to-end Used
  in AAL5
• Cell Loss Priority
• if 1, cell can be discarded by network

GFC-undefined UNI cells has GFC field NNI cells
allocate these 4 bits to VPI 4096 VPs
28
Header Error Check

• The HEC only covers the 5 bytes of the header to
  protect against cell misdelivery
• Since VPI/VCI changes at every switch, HEC must
  be recomputed
• HEC used for cell delineation
• Two modes Header Error Detection / Correction
• Generating Polynomial g(x)x8 x2 x 1
• The pattern 01010101 is XORed to r(x) keeps
  idle cells from having HEC0 and preventing cell
  delineation
• The pattern 01010101 is XORed to r(x) in received
  header prior to error checking

29
ATM Permanent Virtual Connections
Operator at Network Control Center
ATM Switch

• Operator manually sets up VPI/VCI tables at
  switches and terminals
• Long set-up time, long-lived connections

30
ATM Switched Virtual Connections
ATM Switch

• Terminals and switches use pre-defined VPI/VCI to
  setup connections dynamically, on-demand
• Signalling protocol used to communicate with
  call-processing system

31
Traffic Contract

• During connection setup the user and the network
  negotiate two sets of parameters for a connection
• Traffic descriptor the user specifies the
  traffic that it will expect the network to
  transfer on its behalf
• QoS requirements the user specifies the type of
  network performance that is required by its cells
• Traffic Contract
• The user is expected to conform to traffic
  descriptor
• The network is expected to deliver on its QoS
  commitments

32
Quality of Service Parameters

• Six QoS parameters are defined
• Three are intrinsic to network performance and
  are not negotiated during connection setup
• Cell error ratio fraction of delivered cells
  that contain bit errors
• Cell mis-insertion ratio average number of
  cells/second that are misdelivered
• Severely errored cell block ratio M or more out
  of N cells are lost, in error, or misdelivered

33
Negotiable QoS Parameters

• Cell Loss Ratio (CLR) fraction of cells that are
  lost
• Determined by buffer priority
• Cell Transfer Delay (CTD) negotiate maximum
  delay Dmax 1-a of cells have delay less than
  Dmax
• Determined by cell scheduling
• Cell Delay Variation (CDV) Peak-to-Peak
  variation Dmax-D0

34
Traffic Descriptors

• Peak Cell Rate rate in cells/second that a
  source is never allowed to exceed
• Sustainable Cell Rate average cell rate
  produced by the source over a long time interval
  
• Maximum Burst Size maximum number of
  consecutive cells that may be transmitted by a
  source at the peak cell rate (PCR)
• Minimum Cell Rate minimum average cell rate, in
  cells per second, that the source is always
  allowed to send
• Cell Delay Variation Tolerance cell delay
  variation that must be tolerated for in a given
  connection.

35
ATM Service Categories
Cell transfer services provided by ATM Network
VBR real-time
VBR non-real-time
ABR
UBR
CBR
Cell Loss Rate
specified
unspecified
Cell Transfer Delay
specified
unspecified
Cell Delay Variation
specified
unspecified
Traffic Descriptors
PCR/CDVT SCR/BT
PCR/CDVT others
PCR/CDVT
PCR/CDVT
Flow Control
yes
no
no
CBR Constant Bit Rate VBR Variable Bit
Rate ABR Available Bit Rate UBR Unspecified
Bit Rate
PCR Peak Cell Rate CDVT Cell Delay
Variation Tolerance SCR Sustainable Cell
Rate BT Burst Tolerance
36
Multiplexing QoS Guarantees

• ATM provides per-connection QoS guarantees
• Many cell flows are multiplexed onto a common
  stream, so how are guarantees delivered?
• CBR scheduler must ensure transmission
  opportunities are regularly available for each
  connection
• Real-time VBR expect some multiplexing gain
  from combining VBR flows however need to meet
  delay and loss requirements
• Non-real-time VBR can attempt higher
  multiplexing gains, subject only to loss
  requirement
• UBR no guarantees, but excellent performance at
  light traffic
• ABR some degree of guarantee low CLR if
  source responds to network feedback MCR can be
  negotiated

37
Traffic Contract Call Admission Control

• Traffic contract includes the ATM service
  category, the traffic descriptors, and the QoS
  requirements
• Connection admission control (CAC) determines
  whether request for a connection should be
  accepted or rejected
• Each switch in path must determine whether it can
  accommodate new flow and still meet commitments
  to existing flows if yes, resources allocated
• CAC is not standardized, each operator is free to
  select own procedures
• Different degrees of overbooking possible to
  attain different multiplexing gains
• Different types of tariffs for service offerings

38
Policing, Traffic Shaping, and Congestion Control

• QoS guarantees are valid only if the user traffic
  conforms to the connection contract
• Usage parameter control (UPC) is the process of
  enforcing the traffic agreement at the UNI
• Generic Cell Rate Algorithm can be used for UPC
  related to the leaky-bucket algorithm
• Non-conforming cells can be tagged (CLP1) or
  dropped
• Traffic shaping can be used by source to ensure
  that its traffic complies to the connection
  contract
• Token bucket can be used for shaping
• Congestion control
• CLP1 cells are dropped first when congestion
  occurs
• ABR connections must respond to congestion
  feedback information that is received from the
  network
• These topics were discussed in Chapter 7

39
Chapter 9ATM Networks

• ATM Adaptation Layer

40
ATM Adaptation Layer

• AAL end-to-end protocol to adapt the cell
  transfer service provided by ATM network to the
  requirements of specific application classes
• Includes conversion to cells and back, and
  additional adaptation functions, e.g. timing
  recovery, reliable transfer
• ITU defined the following service classes

Class A circuit emulation Class B variable
bit-rate video Class C D packet transmission
41
AAL Protocol Structure

• AAL has two sublayers
• Segmentation Reassembly
• Segments PDUs into cell payloads Reassembles
  PDUs from received cell payloads
• Convergence
• Common Part packet framing and error detection
  functions required by all AAL users
• Specific Part functions that depend on specific
  requirements of AAL user classes

Higher Layers
ATM
42
AAL1

• Provides constant bit rate transfer

43
AAL1

• Convergence Sublayer
• Adaptation to cell-delay variation, constant bit
  rate delivery AAL-SDUs
• Detection of lost or out-of-sequence cells
• Source clock recovery
• Forward error correction on user data
• Forward error correction on Sequence Number (SN)
• 1-bit CS to indicate pointer (used for
  partially-filled cells)
• 3-bit sequence count
• Time-stamp option uses 4 consecutive CS bits for
  residual TS
• SAR Add 1-byte header to 47-byte payload

44
AAL1 services

• Structured Unstructured Transfer
• Unstructured take bits from T1 and group into
  8-bit bytes since T1 frame has 193 bits, bytes
  are never aligned to frame
• Structured take 24 T1 bytes and map into CS
  PDUs use CS PDU pointer to indicate beginning
  of T1 frame
• Forward error control options
• Insert parity cell every 15 cells, correct lost
  cell
• Interleaving of 124 cells, correct up to 4 cell
  losses

45
AAL1 PDUs
SAR PDU header
4 bits
4 bits
46 or 47 octets
Pointer optional
Payload
SN
SNP
8 bits
AAL 1 Pointer
46 Bytes
1 Byte
CS PDU with pointer in structured data transfer
46
AAL2

• New AAL2 intended for bandwidth-efficient
  transfer of low-bit rate, short-packet traffic
  with low-delay requirement
• Adds third level of multiplexing to the VP/VC
  hierarchy of ATM, so low-bit-rate users can share
  an ATM connection.

47
AAL2
48
AAL2 Common Part CS PDU

• Max length CPCS PDU
• 64 bytes
• Channel ID
• Identifies user
• Length Indicator
• Payload length 1
• Packet payload type
• 3 OAM cell
• ?3 application cell
• User-to-user indication
• End-to-end info for application cells
• End-to-end for AAL mgmt when OAM cell
• Error detection
• g(x)x5x21

CPS packet header
49
Packing ATM SDU in AAL2

• CPCS PDUs concatenated, segmented into 48 byte
  chunks, and packed into ATM SDUs
• ATM SDU format

• Offset Field (6 bits)
• From end of the field to start of first CPCS PDU
  or to start of PAD
• Max CPCS PDU may span 2 SDUs
• Sequence Number
• 0 or 1
• Parity bit
• PAD
• 0-47 bytes

50
AAL3/4

• Why 3 / 4 ?
• AAL3 For connection-oriented transfer of data
• AAL4 For connectionless transfer of data
• All connectionless packets use the same VPI/VCI
  at the UNI
• Multiplexing ID (MID) introduced to distinguish
  connectionless packets
• AAL3 and AAL4 combined into AAL that can be used
  for connection-oriented or connectionless
  transfer
• AAL3/4 allows multiple users to be multiplexed
  and interleaved in the same ATM VC
• Message mode single user message segmented into
  ATM payloads
• Stream mode one or more messages segmented into
  ATM payloads and delivered without indication of
  boundaries
• Assured mode error-free delivery of messages
• Non-Assured mode messages may be delivered in
  error, or not at all

51
AAL 3/4
52
AAL3/4 Common Part CS PDU
User Data
Trailer
Header
CPI Btag BASize
AL Etag Length
Pad
CPCS - PDU Payload
1 1 2 1 -
65,535 0-3 1 1 2
(bytes) (bytes) (bytes)

• Common Part Indicator
• How subsequent fields are to be interpreted
• Beginning Tag Ending Tag
• Used to match header trailer at destination

• Buffer Allocation size
• Buffer size required at destination
• Length of payload
• PAD aligns trailer to 32-bit boundary
• Alignment byte of 0s to make trailer 32 bits
  long

53
AAL3/4 SAR PDU

• Segment Type
• 10 Beginning of Message
• 00 Continuation
• 01 End of Message
• 11 Single segment Message
• Sequence Number
• Of SAR PDU within CPCS PDU

• MID allows SAR sublayer multiplexing
• Up to 210 AAL users on 1 ATM VC
• Length Indicator size of payload
• Except for last cell, all cells have LI44
• Last cell has LI 4 to 44
• Each cell payload has 10-bit CRC

54
Multiplexing in AAL3/4
55
AAL3/4 Overhead

• 8 bytes added to each message at CPCS sublayer
• Each ATM payload has 4 out of 48 bytes additional
  overhead
• 9 bytes out of 53 ATM cell bytes overhead
• Too much overhead!
• Let to development of AAL5

56
AAL5
Higher layer
Information

• Simpler than AAL3/4
• 48 bytes payload
• Single packet at a time per VCI
• PTI in ATM header indicates last cell for a given
  packet

Service specific convergence sublayer
Common part convergence sublayer
T
PAD
Information
SAR sublayer

48 (0)
48 (0)
48 (1)

ATM layer
PTI 1
PTI 0
PTI 0
57
AAL5 Common Part CS PDU

• User-to-User 1 byte
• CPI aligns trailer to 8 bytes
• Length 2 bytes to indicate length of CPCS PDU
  payload
• 40-byte CRC

58
Signaling AAL

• AAL standard for BISDN control plane
• Provides reliable transport for signaling
  messages exchanged among endsystems and switches
  to set up ATM VCs.
• SAAL common part a service-specific part
• Service specific part
• service-specific connection-oriented protocol
  (SSCOP)
• Service-Specific Coordination Function (SSCF).
• SSCF supports the signaling applications (UNI and
  NNI).

59
SAAL Process
60
SSCOP PDU

• Padding 0-3 bytes
• Pad Length Indicator
• Reserved (unassigned)
• PDU type
• Sequenced data message poll and control messages
• 24-bit sequence number for large delay-bandwidth
  product
• Depends on error detection provided by AAL5

61
Applications, AALs, and ATM Service Categories

• Applications impose requirements
• Voice, video, connectionless data
• AALs provide segmentation reassembly, and
  possibly additional adaptation functions
• AAL1, AAL2, AAL3/4, AAL5, SAAL
• ATM service category provides cell transfer with
  certain QoS attributes
• CBR, rt-VBR, nrt-VBR, UBR, ABR
• Overall system requirements determine what
  combination of AAL and ATM service category is
  used

62
Application Requirements
63
Summary of AAL Capabilities
64
Examples Voice and Video

• Voice
• AAL1 for individual PCM voice calls
• AAL1 with structured transfer for nx64 kbps
• AAL2 for low-bit-rate cellular voice
• AAL5 for inexpensive voice

• CBR MPEG2 Video
• Timing recovery at AAL or at MPEG systems layer?
• Error detection correction at which layer?
• Timing recovery at MPEG2 systems level and AAL5
  over CBR ATM was selected

65
Example ATM ADSL
Central Office
User Premise
Telephone Switch
Telephone Network
IP PPPoE AAL5 ATM ADSL
Subscriber loop
splitter
ATM Network
ISP
splitter
DSL Access Mux

• IP over PPPoE frames segmented by AAL5 into ATM
  cells at ADSL modem
• ATM cells flow through DSLAM and ATM network to
  Internet Service Provider

66
Chapter 9ATM Networks

• ATM Signaling

67
ATM Signaling

• Signaling means for dynamically setting up and
  releasing virtual connections in ATM
• Signaling involves message exchange across
• User-Network-Interface
• Network-Network Interface
• Broadband Inter-Carrier Interface
• Signaling requires
• Network addressing framework
• Protocols

68
ATM Addressing

• Telephony E-164 Addresses
• For public networks
• Up to 15-digit E-164 (telephone) numbers
• In North America, 1-NPA-NXX-ABCD,
• ATM End-System Addresses (AESAs)
• For private networks
• ISO Network Service Access Point (NSAP) format
• 20 bytes long
• Data Country Code (DCC)
• International Code Designator (ICD)
• E.164 (contained within the AESA format)

69
AESA Address Format
(a) Data Country Code ATM format
1 3

13 19
20
AFI DCC HO-DSP
ESI
SEL
IDP
Domain Specific Part
IDI
(c) E.164 ATM format
1
9
13 19
20
AFI E.164
HO-DSP ESI
SEL
Initial Domain Part
DSP
Initial Domain Identifier
70
ATM Signaling

• Telephone Signaling
• ISDN signaling (Q.931) used in call setup
  messages at the user-network-interface
• Within the network, ISUP protocol of Signaling
  System 7 used to establish a connection from a
  source switch to a destination switch
• For ATM, need UNI, NNI, B-ICI signalling
• UNI Q.2931 ATMF UNI 4.0
• NNI ATMF PNNI based on UNI 4.0
• B-ICI based on B-ISUP

71
UNI 4.0

• ATM connections involve many more parameters than
  narrowband ISDN
• Signaling messages carry Information Elements,
  that describe the user requests
• Signaling messages transferred across the UNI
  using the services of the SAAL layer in the
  control plane
• The signaling cells that are produced by AAL5 use
  the default virtual channel identified by VPI0
  and VCI5.

72
Capabilities of UNI 4.0
73
Signaling Messages
74
UNI Signaling Example
75
PNNI Signaling

• ATM Forum developed PNNI for use
• between private ATM switches (Private Network
  Node Interface)
• between group of private ATM switches (Private
  Network-to-Network Interface)

76
PNNI Protocols

• A routing protocol that provides for the
  selection of routes that can meet QoS
  requirements
• A signaling protocol for the exchange of messages
  between switches and between private networks.
• Based on UNI 4.0 with extensions for
• source routing
• crankback (a feature of the routing protocol)
• alternate routing of connection requests in the
  case of connection setup failure.
• Also includes modifications in the Information
  Elements to carry routing information.

77
PNNI Signaling Example
78
Chapter 9ATM Networks

• PNNI Routing

79
PNNI Routing Protocol

• A routing protocol for the selection of routes
  that can meet QoS requirements
• For intra-domain and inter-domain routing
• Link-state approach each node has network
  topology
• Introduces hierarchy in the ATM network that
  provides a switch
• Detailed routing information in its immediate
  vicinity
• Summary information about distant destinations

80
PNNI Terminology

• Peer Group collection of nodes that maintain an
  identical view of the group
• Logical Group Node abstract representation of a
  peer group at a higher level in the routing
  hierarchy
• Peer Group Leader node in peer group that
  executes functions of LGN for the PG
• Summarizes topology info within the PG
• Injects summary info into higher order groups and
  into the PG

81
PNNI Routing Hierarchy

• PGL passes topology summary upward in hierarchy
  and downwards to its PG
• Multiple levels of hierarchy allowed

82
PNNI Source Routing

• PNNI source node specifies entire path across its
  PG using designated transit list (DTL)
• Rest of path specified using higher levels in the
  hierarchy
• Example station in A.1.1 requests path to B.3
• Path (A.1.1, A.1.2, A.2, B)

83
DTL Stacks Pointers

• DTLs organized in a stack according to level
• A pointer indicates current level

From node B.1 DTL B.1, B.3 pointer-2 DTL A,
B pointer-1
84
PNNI Features

• Call setup involves connection admission control
  at each node
• PNNI uses Generic Connection Admission Control
  (GCAC) to select path
• Call request can be blocked from lack of
  resources
• PNNI provides for crankback alternate routing
• Upon blocking, call setup is cranked back to
  creator of DTL, which considers alternate routes
  from that point onwards

85
Chapter 9ATM Networks

• Classical IP over ATM

86
Classical IP over ATM

• Classical IP over ATM (RFC 2255)
• IP treats ATM as subnetwork
• Logical IP subnetwork (LIS) is part of ATM
  network that belongs to same IP subnetwork
• All members of a LIS use same IP address prefix
  (network subnetwork )
• Members in same LIS communicate using ATM VC
• Each LIS in an ATM network operates independently
  of other LISs in the same ATM network
• LISs communicate via routers

87
Logical IP Subnetworks (LISs)
88
Address Resolution

• Suppose host S want to send packet to host D in
  same LIS
• Host S sends message to ATM ARP server in the
  LIS, requesting ATM address corresponding to IP
  address of host D
• (All hosts in LIS know ATM address of ATM ARP
  server)
• ATM ARP replies with ATM address, and Host S sets
  up ATM connection to Host D
• If host D is in another LIS, host S sets up ATM
  connection to the router in its LIS
• Router determines next hop router sets up VC to
  it
• Packets between hosts in different LISs always
  use intermediate routers, even if hosts are in
  the same ATM network