Transcript ppt - Dr. Wissam Fawaz

Chapter 3:
ATM Networks
TOPICS
–
–
–
–
The ATM header
The ATM protocol stack
The physical layer
ATM switch architectures
– ATM adaptation layers
– IP over ATM
Connection-Oriented Networks – Wissam Fawaz
1
Asynchronous Transfer Mode (ATM)
• The word Asynchronous in ATM is in
contrast to Synchronous Transfer Mode
(STM) that was proposed earlier on, which
was based on the SONET/SDH hierarchy.
• Transfer Mode refers to a telecommunication
technique
Connection-Oriented Networks – Wissam Fawaz
2
• ATM was standardized by ITU-T (old CCITT) in
1988 as the transfer mode of B-ISDN
• It can carry a variety of different types of traffic,
such as
– Voice
– Video
– Data
At speeds varying from fractional T1 to 2.4 Gbps
Connection-Oriented Networks – Wissam Fawaz
3
• These different types of traffic have
different Quality-of-Service (QoS)
requirements, such as:
– Packet loss
– End-to-end delay
• ATM, unlike IP networks, can provide each
traffic connection a different type of quality
of service.
Connection-Oriented Networks – Wissam Fawaz
4
Some features of ATM
• Connection-oriented packet-switched network
• Fixed cell (packet) size of 48+5 bytes
Header
Payload
5 bytes
48 bytes
• No error protection on a link-by-link
• No flow control on a link-by-link
• Delivers cells in the order in which they were
transmitted
Connection-Oriented Networks – Wissam Fawaz
5
The structure of the ATM cell
NNI cell format
UNI cell format
1
1
2
2
3
4
5
.
.
.
5
6
GFC
VPI
VPI
VCI
3
B
y
t
e
4
7
8
2
PTI
2
3
CLP
HEC
Information
payload
53
Connection-Oriented Networks – Wissam Fawaz
B 4
y
t
e 5
.
.
.
4
5
6
7
8
VPI
VCI
VPI
3
VCI
VCI
1
1
VCI
VCI
PTI
CLP
HEC
Information
payload
53
6
Fields in the ATM cell header
•
•
•
•
•
GFC
Connection identifier: VPI/VCI,
Payload type indicator (PTI)
Cell loss priority (CLP)
Head error control (HEC)
Connection-Oriented Networks – Wissam Fawaz
7
ATM connections
• Identified by the combined fields
– virtual path identification (VPI), and
– virtual channel identification (VCI)
• VPI field:
– 256 virtual paths at the UNI interface, and
– 4096 virtual paths at the NNI interface.
• VCI field:
– a maximum of 65,536 VCIs.
Connection-Oriented Networks – Wissam Fawaz
8
• VPI/VCI values have local significance.
That is, they are only valid for a single hop.
• A connection over many hops, is associated
with a different VPI/VCI value on each hop.
• Each switch maintains a switching table.
For each connection, it keeps the incoming
and outgoing VPI/VCI values and the input
and output ports.
Connection-Oriented Networks – Wissam Fawaz
9
Label swapping
C
A
VPI=30
VCI=41
1
VPI=40
VCI=62
2
ATM
switch 1
30 41 1 30 53 4
40 62 2 10 89 3
4
3
VPI=30
VCI=53
4
ATM
switch 2
VPI=10
VCI=89
VPI=100
VCI=53
5
D
30 53 4 100 53 5
1
ATM
switch 3
10 89 1 50 77 6
6
VPI=50
VCI=77
B
Connection-Oriented Networks – Wissam Fawaz
10
PVCs and SVCs
• Depending how a connection is set-up, it
may be
– Permanent virtual circuit (PVC)
– Switched Virtual circuit (SVC)
• PVCs are set-up administratively. They
remain up for a long time.
• SVCs are set-up in real-time using ATM
signalling. Their duration is arbitrary.
Connection-Oriented Networks – Wissam Fawaz
11
Payload type Indicator
•
•
•
•
•
•
•
•
•
PTI
000
001
010
011
100
101
110
111
Meaning
User data cell, congestion not experienced, SDU type=0
User data cell, congestion not experienced, SDU type=1
User data cell, congestion experienced, SDU type=0
User data cell, congestion experienced, SDU type=1
Segment OAM flow-related cell
End-to-end OAM flow-related cell
RM cell
Reserved
Connection-Oriented Networks – Wissam Fawaz
12
Head Error Control (HEC)
Multiple bit error detected
(cell discarded
No error detected
(No action)
Correction
mode
No error detected
Detection
mode
Error detected
(cell discard)
Single bit error detected
(correction)
Connection-Oriented Networks – Wissam Fawaz
13
The ATM protocol stack
voice
Video
Data
ATM adaptation layer
ATM layer
Physical layer
Connection-Oriented Networks – Wissam Fawaz
14
The physical layer
• The physical layer transports ATM cells
between two adjacent ATM layers.
• It is subdivided into
– transmission convergence (TC) sublayer
– physical medium-dependent (PMD) sublayer.
Connection-Oriented Networks – Wissam Fawaz
15
The transmission convergence (TC)
sublayer
• HEC cell generation and verification
– Implements the HEC state machine
• Decoupling of cell rate
– Maintains a continuous bit stream by inserting idle cells
• Transmission frame generation and recovery
– Such as SONET frames
• Cell delineation
Connection-Oriented Networks – Wissam Fawaz
16
Cell delineation is the extraction of cells from the
bit stream received from the PMD sublayer.
hunt
Incorrect
HEC for a cells
Sync
Incorrect
HEC
Correct
HEC
Correct
HEC for d cells
Presync
Connection-Oriented Networks – Wissam Fawaz
17
• Physical medium dependent (PMD)
– Timing function
• Used to synchronize the transmitting and receiving
PMD sublayers.
– Encoding/decoding
• PMD may operate on a bit-by-bit basis or using
block coding such as 4B/5B and 8B/10B schemes.
Connection-Oriented Networks – Wissam Fawaz
18
ATM physical layer interfaces
•
•
•
•
•
•
SONET/SDH
Plesiochronous digital hierarchy (PDH)
Nx64 Kbps
Inverse mulitplexing for ATM (IMA)
asymmetric digital subscriber line (ADSL)
APON
Connection-Oriented Networks – Wissam Fawaz
19
The ATM layer
• The ATM layer is concerned with the endto-end transfer of information, i.e., from the
transmitting end-device to the receiving
end-device.
• Below, we summarize its main features.
Connection-Oriented Networks – Wissam Fawaz
20
Connection-oriented packet switching
• The ATM layer is a connection-oriented point-to
point packet-switched network with fixed-size
packets (known as cells).
• A connection is identified by a series of VPI/VCI
labels, as explained above, and it may be point-topoint or point-to-multipoint.
• Cells are delivered to the destination in the order
in which they were transmitted.
Connection-Oriented Networks – Wissam Fawaz
21
Cell switching in ATM networks
is carried out at the ATM level
Connection-Oriented Networks – Wissam Fawaz
22
No error and flow control
on each hop
• Low probability of a cell getting lost or delivered to the
destination end-device in error.
• The recovery of the data carried by lost or corrupted cells
is expected to be carried out by a higher-level protocol,
such as TCP.
• When TCP/IP runs over ATM, the loss or corruption of the
payload of a single cell results in the retransmission of an
entire TCP PDU.
Connection-Oriented Networks – Wissam Fawaz
23
Addressing
• Each ATM end-device and ATM switch has a unique
ATM address.
• Private and public networks use different ATM
addresses. Public networks use E.164 addresses and
private networks use the OSI NSAP format.
• ATM addresses are different to IP addresses.
Connection-Oriented Networks – Wissam Fawaz
24
Quality of service
• Each ATM connection is associated with a
quality-of-service category.
• Each quality-of-service category is associated with
a set of traffic parameters and a set of quality-ofservice parameters.
• The ATM network guarantees the negotiated
quality-of-service for each connection.
Connection-Oriented Networks – Wissam Fawaz
25
Congestion control
• In ATM networks, congestion control
permits the network operator to carry as
much traffic as possible without affecting
the quality of service requested by the users.
• It consists of call admission control and a
policing mechanism.
Connection-Oriented Networks – Wissam Fawaz
26
The ATM switch architecture
CPU
Input
queues
1
1
...
...
Incoming
links
Output
queues
N
Outgoing
links
N
Switch fabric
A generic ATM switch
Connection-Oriented Networks – Wissam Fawaz
27
The shared memory ATM switch
architecture
S hared memory
1
1
...
...
N
Connection-Oriented Networks – Wissam Fawaz
N
28
• If the speed of transmission on each
incoming and outgoing link is V, then the
switch can keep up at maximum arrival rate,
if the memory's bandwidth is 2NV
• Total memory capacity is B cells
• Each linked list i is associated with a
minimum dedicated space and it is limited
to an upper bound Bi, Bi<B so that SBi>B.
Connection-Oriented Networks – Wissam Fawaz
29
Cell loss in a shared memory switch
• Cell loss occurs when a cell arrives at a time when
the shared memory is full, that is it, contains B
cells.
• Cell loss can also occur when a cell with
destination output port i arrives at a time when the
total number of cells queued for this output port is
Bi cells (even if the total number of cells in the
shared memory is less than B.)
Connection-Oriented Networks – Wissam Fawaz
30
Non-blocking output buffering switch
• In a non-blocking switch, the switching fabric
does not give rise to internal or external blocking.
• An output buffering switch has buffers only at its
output ports.
...
...
output ports
• A shared memory switch is non-blocking with
output buffering
Connection-Oriented Networks – Wissam Fawaz
31
Scheduling algorithms
• Let us consider a non-blocking switch with output
buffering. Each output buffer holds cells that
belong to different connections.
• Each of these connections is associated with a
quality-of-service category.
• The cells belonging to these connections are
grouped into queues, one per quality-of-service
category, and these queues are served using a
scheduling algorithm.
Connection-Oriented Networks – Wissam Fawaz
32
Static priorities
CBR
RT-VBR
From
switch
fabric
NRT-VBR
Output
port
ABR
UBR
Connection-Oriented Networks – Wissam Fawaz
• Priorities among the
queues
• Always serve high
priority queue first,
then second priority
queue, etc.
• Aging
• Purging
33
Early deadline first (EDF) algorithm
• Each cell is assigned a deadline upon arrival at
the buffer. The scheduler servers the cells
according to their deadlines, so that the one with
the earliest deadline gets served first.
• Cells belonging to delay-sensitive applications,
such as voice or video, can be served first by
assigning them deadlines closer to their arrival
times.
Connection-Oriented Networks – Wissam Fawaz
34
The round-robin scheduler
• Each output buffer is organized into a number of
logical queues.
• The scheduler serves one cell from each queue in
a round robin fashion
• Empty queues are skipped
• Weighted round robin scheduling can be used to
serve a different number of cells from each queue
Connection-Oriented Networks – Wissam Fawaz
35
The ATM adaptation layer
• The purpose of AAL is to isolate higher
layers from the specific characteristics of
the ATM layer.
• AAL consists of the
– convergence sublayer, and the
– segmentation-and-reassembly sublayer.
Connection-Oriented Networks – Wissam Fawaz
36
The AAL sublayers
SAP
Convergence
Sublayer
Service Specific
Convergence Sublayer (SSCS)
Common Part
Sublayer (CPS)
Segmentation and Reassembly
SAP
Connection-Oriented Networks – Wissam Fawaz
37
ATM Adaptation Layer 1 (AAL 1)
• This AAL can be used for applications such as:
– circuit emulation services
• It emulates a point-to-point TDM circuit over ATM
– Constant-bit rate audio
• Used to provide an interconnection between two
PBXs over a private or public ATM network
Connection-Oriented Networks – Wissam Fawaz
38
The SAR encapsulation for AAL 1
payload
SAR Header
SN
CSI Sequence. count
1 bit
3 bits
SNP
47 bytes
CRC-3
3 bits
Connection-Oriented Networks – Wissam Fawaz
Parity
1 bit
39
The AAL 1 CS functions:
1. Handling of cell variation
–
Due to queueing delays, inter-arrival times of cells vary (jitter).
Sender
cell
i
cell
i-1
Receiver
cell
i+1
ATM
cloud
t
t
i-1
i
Inter-departure gaps
–
cell
i-1
cell
i
cell
i+1
si-1
si
Inter-arrival gaps
CS writes received data into a buffer, and then delivers the
information to the application at constant bit rate.
Connection-Oriented Networks – Wissam Fawaz
40
2. Processing of the sequence count
– The sequence count values are processed by CS in
order to detect lost or misinserted cells. Detected
misinserted cells are discarded. In order to maintain
bit count integrity of the AAL user information, it
may be necessary to compensate for lost cells by
inserting dummy SAR-PDU payloads.
3. Forward error correction
– For video and high quality audio forward error
correction may be performed in order to protect
against bit errors. This may be combined with
interleaving of AAL user bits to give a more secure
protection against errors.
Connection-Oriented Networks – Wissam Fawaz
41
4. Transfer of timing information
a.
Synchronous residual time stamp (SRTS):
CS conveys to the receiver in the CSI field the
difference between a common clock derived from the
network and the sender’s clock
b. Adaptive clock method:
The receiver writes the received information into a
buffer and reads out from the buffer. If its clock is
fast/slow the occupancy in the buffer will be
below/over the median
Connection-Oriented Networks – Wissam Fawaz
42
5. Structured and unstructured data transfers
Two CS-PDU formats have been defined:
a. CS-PDU non-P format:
Constructed from 47 bytes of information
supplied by an AAL user
b. CS-PDU P format:
Constructed from a 1-byte header and 46 bytes
of information supplied by an AAL use.
The header consists of a 7-bit pointer (SDT
pointer) and 1 even bit parity.
Connection-Oriented Networks – Wissam Fawaz
43
Circuit Emulation Services
UNI
CBR
User A
•
•
IWF
A
UNI
ATM network
IWF
B
CBR
User B
The structured and unstructured data transfers are
used in Circuit Emulation Services (CES), which
emulate a T1/E1 link over ATM.
CES is implemented in an interworking function
(IWF).
Connection-Oriented Networks – Wissam Fawaz
44
• Unstructured service
– The entire DS1/E1 signal is transported by
packing it bit by bit into the 47-byte payload of
a CS-PDU non-P format, which is then carried
by an ATM cell.
47 bytes -> 376 bits -> less than 2 DS-1
frames (193 bits/frame)
Connection-Oriented Networks – Wissam Fawaz
45
• Structured transfers
– It is used to carry fractional T1/E1, i.e. Nx64
Kbps
– Fractional T1/E1 generates blocks of N bytes
every 125 msec. Such a block of data is referred
to in the standards as a structured block.
– Blocks of N bytes are transported back-to-back
over successive cells using both the CS-PDU
non-P and P formats.
Connection-Oriented Networks – Wissam Fawaz
46
• The SDT pointer
– The SDT pointer in the CS-PDU P format is
used to help delineate the boundaries of these
blocks.
– The actual rules as to when to use the SDT
pointer in the P format are somewhat complex.
Connection-Oriented Networks – Wissam Fawaz
47
An example: block size = 20 bytes
Seq, count 0
20
Seq, count 1
20
6
Seq, count 4
13
20
13
7
Seq, count 5
20
14
Seq, count 0
6
14
Seq, count 2
6
20
20
20
20
20
1
19
7
Connection-Oriented Networks – Wissam Fawaz
13
20
7
20
14
Seq, count 7
20
13
7
Seq, count 2
20
20
20
Seq, count 6
Seq, count 1
20
Seq, count 3
Seq, count 3
20
14
6
20
20
1
48
ATM Adaptation Layer 2 (AAL 2)
• Defined for delay sensitive applications
with a low bit rate, such as voice and
voiceband traffic (facsimile, modem traffic)
• AAL 2 is used to interconnect two distant
public or private telephone networks over
an ATM network.
Connection-Oriented Networks – Wissam Fawaz
49
• At the sender, AAL 2 multiplexes several streams
onto the same ATM connection
• At the receiver, it de-multiplexes the date from
the connection to the individual streams.
Connection-Oriented Networks – Wissam Fawaz
50
F
Time Time Time
slot 1 slot 2 slot 3
PBX A
...
Time
slot 24
T1/E1
IWF
IWF
T1/E1
PBX C
ATM
network
PBX B
T1/E1
IWF
IWF
T1/E1
F
Connection-Oriented Networks – Wissam Fawaz
Central
office
Time Time Time
slot 1 slot 2 slot 3
...
Time
slot 24
51
The SSCS and CPS sublayers
• The AAL 2 services are provided by the
convergence sublayer, which is subdivided into
the
– Service Specific Convergence Sublayer (SSCS)
– Common part sublayer (CPS).
Connection-Oriented Networks – Wissam Fawaz
52
Functional model of AAL 2 (sender side)
• Each stream is served by a separate SSCS which is
associated with a CID
AAL-SAP
SSCS
SSCS
CID=X
SSCS
CID=Z
CID=Y
CPS
ATM-SAP
Connection-Oriented Networks – Wissam Fawaz
53
SSCS for AAL 2 trunking
• A specialized SSCS has been developed to
support “ATM trunking using AAL 2 for
narrowband services”.
• It is described in Chapter 12
Connection-Oriented Networks – Wissam Fawaz
54
CPS-packets and CPS-PDUs
• A transmitting SSCS uses a timer to decide
when to pass on the data to CPS.
• Data from an SSCS is packed into a CPS-packet
• CPS-packets from different SSCSs are packed
into a CPS-PDU, which is exactly 48 bytes and
it is carried in an ATM cell
Connection-Oriented Networks – Wissam Fawaz
55
Packing CPS-packets into CPS-PDUs
CPS-packets
CPS-PDUs
1
2
1
3
2
3
4
3
5
4
5
ATM cells
Connection-Oriented Networks – Wissam Fawaz
56
The structure of the CPS-packet
and CPS-PDU
1
2
3 4
3
6 7 8
CID
1
2
5
PPT
1
1
2
P S
N
3 4
5
6 7 8
OSF
LI
HEC
UUI
CPS-PDU
payload
CPS-packet
payload
48
CPS-packet
Connection-Oriented Networks – Wissam Fawaz
PAD
CPS-PDU
57
The CPS-packet fields
• Channel identifier (CID) - 8 bits:
– Identifies a channel. Same value is used for both directions.
– CIDs are allocated using the AAL negotiation procedures (ANP)
• Packet payload type (PPT) - 2 bits:
– Indicates whether it carries voice or network management data
• Length indicator (LI) - 6 bits:
– Default maximum length of the CPS-packet payload is 45 bytes.
• Header error control (HEC) - 5 bits:
– Pattern is: x5+x2+1.
• User-to-user-indication (UUI) - 3 bits:
– Used to transfer information transparently between the peers.
Connection-Oriented Networks – Wissam Fawaz
58
CPS-PDU fields
Offset field (OSF) (6 bits)
• Used to identify the beginning of a CPSpacket. It points to the first new CPS-packet
in the CPS-PDU payload
• In the absence of a new CPS-packet, it
points to the beginning of the pad
• The value of 47 indicates that there is no
beginning of a CPS-packet in the CPS-PDU.
Connection-Oriented Networks – Wissam Fawaz
59
An example
20
48
20
#1
27
35
21
#2
#2
OSF =0
26
#3
20
9 20
#3 #4
padding
OSF=21
Connection-Oriented Networks – Wissam Fawaz
OSF=9
60
ATM adaptation layer 5 (AAL 5)
• Very popular AAL due to its simplicity
• A user-PDU is encapsulated and then broken up to
fragments, each carried by an ATM cell
• AAL 5 consists of
– Convergence sublayer (CS)
• SSCS
• CPS
– Segmentation and reassembly (SAR).
Connection-Oriented Networks – Wissam Fawaz
61
CPS
• Provides a non-assured transfer operation.
• User-PDUs of a length up to 65,535 bytes
can be transferred.
• Erroneous CPS-PDUs can be detected at the
receiver’s side. No recovery of an erroneous
CS-PDU takes place. Instead, an indication
is sent to the the higher-level application.
Connection-Oriented Networks – Wissam Fawaz
62
CPS encapsulation
• Pad: from 0 to 47 bytes, so that the entire CPS-PDU becomes
an integer multiple of 48 bytes. The User-PDU can be up to
65,535 bytes
• CPS User-to-user indication (CPS-UU): 1-byte field
• Common part indicator (CPI): 1-byte field for future use
• Length: 2-byte field gives length of User-PDU.
• CRC pattern: 4-byte field contains the FCS calculated using
the pattern x32+x26+x23+x22+x16+x12+x11+x10+x8+x7+x5+x4+x2+x+1.
Connection-Oriented Networks – Wissam Fawaz
63
SAR sublayer - transmitter
• SAR segments a CPS-PDU into a sequence
of 48-byte segments.
• No additional encapsulation
• Each segment is carried in the payload of an
ATM cell
• Last cell has its PTI marked with SDU=1.
Connection-Oriented Networks – Wissam Fawaz
64
SAR sublayer - Receiver
• SAR appends payloads of the ATM cell
into a buffer until
1. It encounters an SDU=1 in PTI field
• It checks the CRC and then passes the PDU to the
application above with an indication as to whether
it is correct or not.
2. Buffer is exceeded
• It passed the PDU to the application above with an
indication that buffer was exceeded.
Connection-Oriented Networks – Wissam Fawaz
65
Classical IP and ARP over ATM
• This is a technique proposed by IETF for
supporting IP over ATM in a single logical IP
subnet (LIS).
• A LIS is a group of IP hosts that share a common
IP network address and subnet mask, and they all
communicate with each other directly over ATM
connections.
Connection-Oriented Networks – Wissam Fawaz
66
A logical IP subnet (LIS)
Group of host with the same network address
and same subnet mask
IP address: 193.14.0.0
Connection-Oriented Networks – Wissam Fawaz
We now replace the transport network
with ATM switches
67
Classical IP
Computer
Computer
TCP
IP
TCP
IP
CIP
AAL
ATM
PHY
CIP
AAL
ATM
PHY
ATM switch
IP packet
CS-PDU
IP packet
CS-PDU
SAR
SAR
ATM
ATM
Connection-Oriented Networks – Wissam Fawaz
68
ATMARP and InATMARP
• A LIS member has both an IP and an ATM
address.
• IP addresses are resolved to ATM addresses using
the ATMARP protocol within the LIS (based on
ARP).
• The inverse ATMARP (InATMARP) protocol is
used to resolve an ATM address to an IP address.
(It is similar to the inverseARP).
Connection-Oriented Networks – Wissam Fawaz
69
The ATMARP server
• It maintains a table or a cache of IP and ATM
address mappings
• It learns about the IP and ATM addresses of the
LIS members (IP clients) through the messages
exchanged between ATMARP and the LIS hosts.
• It typically resides on an ATM switch. (An ATM
switch load commonly contains the ATMARP
server as well).
Connection-Oriented Networks – Wissam Fawaz
70
ATMARP messages
• ATMARP_request: Used to request the ATMARP server
the ATM address of a destination IP client.
• ATMARP_reply: Used by the ATMARP server to respond
to an IP client with the destination ATM address.
• InATMARP_request: Sent from the ATMARP server to an
IP client to obtain its IP address.
• InATMARP_reply: This is the response from an IP client
with its IP address.
• ATMARP_NAK: Negative response issued by the
ATMARP server to a requesting IP client.
Connection-Oriented Networks – Wissam Fawaz
71
Registration
• An IP client must first register its IP and ATM
addresses with the ATMARP server.
– The client establishes a connection to the ATMARP
server (it knows the ATMARP server’s ATM address).
– It then transmits an ATMARP_request, asking its own
ATM address.
– The ATMARP server checks against duplicate entries
in its table, time stamps the entry, adds it to its table,
and sends an ATMARP_reply. The entry is valid for a
minimum of 20 minutes and has to be refreshed.
Connection-Oriented Networks – Wissam Fawaz
72
Address resolution
• An ATMARP client 1 can communicate with an
ATMARP client 2 immediately if it knows its
ATM address.
• If the destination ATM address is not known,
client 1 invokes the ATMARP process.
– It sends an ATMARP_request to the ATMARP server.
– If the server has the requested address in its table, it
returns an ATMARP_reply.
– Otherwise, it returns an ATMARP_NAK.
Connection-Oriented Networks – Wissam Fawaz
73