Transcript Router

Lecture 2
Advanced Networking CSE 8344
Southern Methodist University
Fall 2003
Mark E. Allen
Introduction
• Miscellaneous notes:
– Exam dates: (posted on web also)
SONET
• SONET byte oriented frame format
– Path, line, and section
– Multiplexing format
• Virtual Containers
• Synchronous payload envelope (SPE)
• Pointers
– Timing issues
– This is what makes SONET synchronous -- the payload can float
in the SONET frame.
• Overhead
– Line, Section, and Path
• Performance monitoring
SONET Hierarchy
Signal
Bit rate
Capacity
STS-1, OC-1
51.840 Mbps
28 DS1s, 1 DS3
STS-3, OC-3
155.520 Mbps
84 DS1s, 3 DS3s
STS-12, OC-12
622.080 Mbps
336 DS1s, 12 DS3s
STS-48, OC-48
2488.320 Mbps
1344 DS1s, 48 DS3s
STS-192, OC-192
9953.280 Mbps
5376 DS1s, 192 DS3s
SONET Network
LINE
SECTION
Terminal
(LTE)
REG
ADM
or
DCS
(LTE)
REG
REG
Terminal
(LTE)
Router
(PTE)
Router
(PTE)
PATH
SONET Networking
810 bytes x 8000 frame/sec x 8 bits = 51,840,000 bps
OH
PAYLOAD
9 rows
OH
PAYLOAD
OH
STS-1
Synchronous
Payload
Envelope
90 columns (87 columns of payload)
3 columns of
transport overhead:
Section overhead
Line overhead
Path overhead
PAYLOAD
BIP-8/BI: Parity Checking
9 Rows
STS-1 Frame Section Trace/Growth
A1
A2
JO/Z0
J1
B1
E1
F1
B3
D1
D2
D3
C2
H1
H2
H3
G1
B2
K1
K2
F2
D4
D5
D6
H4
D7
D8
D9
Z3
D10
D11
D12
Z4
E2
Z5
S/Z1 M0/M1
87 Columns of Payload
STS-1 Synchronous
Payload Envelope
(STS-1 SPE)
BIP-8/B2: Error Monitoring
Z2
90 Columns
SONET Overhead
• Path overhead – Overhead for entire end-to-end
circuit. Path terminating equipment (PTE)
terminates this overhead.
• Line overhead – Overhead for connection between
terminals, cross-connects (DCS), or add-drop
mulitplexer (ADM). Line terminating equipment
(LTE) terminates this overhead.
• Section overhead – Overhead for connection
between regenerators. Section terminating
equipment (STE) terminates this overhead.
Path overhead
J1
B3
C2
G1
F2
H4
Z3
Z4
Z5
• J1 – Used for tracing the circuit
• B3 – Bit interleaved parity byte to check for
errors
• C2 – Path signal label byte to indicate the
contents of the SPE
• H4 – Virtual tributary (VT) multiframe
indicator byte to describe multiframe VT
payloads (pointer)
• G1 – Path status byte so PTE can detect
problems on the path
• F2 – Path user channel byte for
communications between path elements
• Z3,Z4,Z5 – User bytes reserved for future
Line
overhead
• H1,H2 – pointer to beginning of SPE
H1
H2
H3
B2
K1
K2
D4
D5
D6
D7
D8
D9
D10
D11
D12
S1/Z1
M0 or
M1/Z
2
E2
• H3 – pointer action byte used to hold
data when pointer adjustment is made
• B2 – byte interleaved parity for line
• K1, K2 – used to manage Auto-Protection
Switching (APS)
• D4 to D12 – Data communication
channel (DCC) bytes are 576 kbps
• Z1, Z2 – Not defined (Z2 used for FEBE
in STS-3)
• E2 – express 64 kbps channel between
LTE (for STS-1 only)
Section overhead
A1
A2
C1
B1
E1
F1
D1
D2
D3
• A1,A2 – Framing bytes
• C1 – STS ID set for each
STS1
• B1 – Byte interleaved parity
for monitoring for errors
• E1 – 64 kbps orderwire
• F1 – Used by section
equipment
• D1 to D3 – Data
communication channel
SONET multiplexing
• To create OC-N signals, SONET streams
are BYTE INTERLEAVED.
• No bit-stuffing is used because network is
synchronous
• Pointers are used to account for phase
differences in SPE of tributary signals
(tribs).
• Overhead from all tribs is aligned.
Typical point to point SONET
link
Atlanta
Dallas
OC48
OC48
OC48
OC48
O W
C
1
9 P
2
O
C
1
9
2
40 mile amplifier spacing
Traditional physical layer
switching
Purpose of switches
• Voice networks
– Connect dialing party to called party. Required making
connections at CO, Tandem, IXC switching nodes
• Layer 1 data network
– Connection at SONET or optical layer to connect a DS3
or OC3 through the network
• Layer 2 data network
– Connections via ATM or Ethernet switch to establish a
flow, or PVC through which two PCs or routers are
connected
Switching systems
• Switches vs. Routers
– Switches are typically connection oriented, routers are
based on datagram routing
• Routers use routing lookup tables to send out the packet.
• Switches are based on a connection, flow, or circuit that
usually traverses several switches from source to destination
• Packet switches have queues, circuit switches do not.
• Focus on circuit switching for now
– Packet switching and routing are an extension of circuit
switching.
– Optical switching is simpler (conceptually not
technically) subset, to be discussed later.
Basic crosspoint switch
(Bellamy)
Space switch
• Most basic switch, sometimes called crosspoint
switch.
– Rectangular fabric: any input can connect to any output
– Number of crosspoints is N X M
• Graded switches
– Each input has access to a select group of outputs
– Used when crosspoints are expensive or switch would
be too big.
Graded matrix (Bellamy)
Switches (cont)
• Square vs. triangular
– Square fabrics have two possible ways of making
connection
– Triangular get rid of extra cross points but require
compare
• Why multistage switches
– For square fabric, N(N-1) switches required
– For triangular array, N(N-1)/2 required
– This results in too many pieces for a practical sized
switch, 5 Billion crosspoints for 100,000 port switch.
– Multistage switching is the answer
Three-stage (Bellamy)
3 stage switches
• Number of cross points in 3 stage switch is:
• Nx=2Nk + k (N/n)^2
– Where N is number of inputs
– k is number of center stages
– n is size of inlet / outlet group
• Consider what happens with blocking
– There is no center stage that can make a connection to
output stage that can switch to the desired output.
• Clos showed that if
– k = (n-1) + (n-1) + 1 then switch is non-blocking
3 stage switch (cont)
• Using this k for number of crosspoints
yields (equ 5.2, Bellamy)
•
• Solving for the minimum number of
crosspoints yields (equ 5.3)
3-stage (cont, Bellamy)
• Note the reduction in required cross points
in (Table 5.1) by using a 3 stage Clos
switch.
Crosspoint reduction (Bellamy)
Blocking switches
• In reality, Clos switches are “rearrangeably”
non blocking. Not strictly non-blocking.
• In real-world, connections are continuously
being made and torn down.
– So we can’t pick the perfect path for each
connection beforehand.
– Clos switch still requires a fair number of
crosspoints.
Switches with blocking
• It’s often practical to make a switch that is
“blocking”
– There is some small probability the switch can’t
connect an input to an output
– Recall it depends on what other connections
have been made (i.e. how “busy” is the switch?)
– Many switches aren’t very busy
– Considerable cost savings can be enjoyed by
reduction in cross points
Analysis of blocking switches
• These equations provide probability blocking
through a switch fabric
– Lee graphs
– Jacobaeus
• Discrete event simulation software packages are
often used in practice when designing switches
– Modeling input behavior is a challenge
• Call times, relationship between inputs and outputs, etc.
– Examples: OPNET
Time Division switching
• Time Division Switching allows multiple
connections to share cross-points
– Results in even fewer cross points than
multistage switches
• Goes well with Time division multiplexing
– Many times, the individual circuits have been
TDM’d prior to being connected to the switch
Time slot interchange (TSI)
• This is an important function of digital switches.
• Memory is used to rearrange data in the time slots
• Allows information to arrive at the Space switches
at the right time.
• Normally used with Space switching to create
TST, TSST, etc. matrices that combine both.
• The Lucent 4ESS switch uses TSSSST.
– 4 internal Switches wrapped in two TSI switches
– Can handle from 100,000 to 200,000 calls.
Keshav, Chapter 8
Keshav, Chapter 8
Bellamy Ch 5
Cross-connects
• Digital Cross-connect is a specialized switch
fabric
– Combines muxing and switching
– Used to aggregate (fill) and groom
• Typically appear as 3/3, 3/1, 3/1/0, etc.
–
–
–
–
3/3 cross connects DS3s
3/1 cross connects DS1 within DS3s or entire DS3.
3/1/0 groom to the DS0 level.
Lower granularity of grooming costs more (more
crosspoints)
– Hierarchy is often used (see figure 5.34)
Big switch example
• Example:
– SONET cross-connects
– 256 OC48 external interfaces
– What is total switch capacity?
• 256 X 2.5 Gbps = 640 Gbps
– IF DS0s were to be groomed, how many possible
connections?
• 256 X 48 X 28 X 24 = 8.25 Million input channels
• Using N(N-1)/2, would be huge!, even Clos is too big
– What about STS-1 granularity?
• Switches exist to do this (barely!)
Voice Network Signaling
Voice network signaling
• Signaling function
– Supervisory
• On hook, Off hook, dial tone, ringing, on-hook, busy signal
– Information bearing
• Dialed digits, toll charges, etc.
– In-channel signaling
• In band
– Single Frequency (SF), dual tone multifrequency (DTMF),
multifrequency (MF) which all operate in voice band
• Out of band
– DC levels on the loop portion or out of band using FDM
– Pulses on phones for dialed digits are out of band
Signaling (cont)
• Common channel signaling (CCS)
– Here, the signaling information is contained in a
separate signaling channel
– Channel is carries signaling for several lines
– Good for fraud prevention
– Simpler to manage signaling between switches
– Disadvantages:
• Signaling may not propagate through the network to free
resources
• No automatic testing
• Trunks may not all terminate at the same switch (signaling
must be forwarded)
Analog interfaces
•
Subscriber loop interfaces
1. Battery: 48 volts is supplied to operate the phone
2. Overvoltage protection: protection from lightning,
etc.
3. Ringing: 20Hz 86 volt rms signal to ring the phone.
4. Supervision: Detection of on/off hook
5. Test: Access to testing the loop
BORSCHT (Battery, Overvoltage, ….)
(Hybrid and Coding are also required at switch end.)
Analog interfaces (cont)
• Loop start trunks
– Simple connection between switches
• Central Office to Private Branch Exchange (PBX)
• Problem of “glare” exists
• Ground start trunk solve this problem
– More elaborate communication between PBX and CO
• Direct Inward Dial (DID) trunks
– Allow incoming calls to PBX to connect directly to
called party (no attendant necessary)
Analog interfaces (cont)
• E&M (ear and mouth) trunks
–
–
–
–
5 types of E&M interfaces are defined
Type II E&M is a 8 wire interface
TX pair, RX pair, E pair, and M pair
Supervisory signaling happens over the E&M
leads
– Typically used to connect PBX to CO, PBX to
PBX
Digital Networking
• Analog loops will exist for some time
– Businesses will move to digital phones more quickly
• For switching and transport, analog has serious drawbacks
– Noise, Ease of multiplexing, switching,
• Current approach is to convert to digital at the ingress of
the network
–
–
–
–
–
Digitization schemes will be discussed later
Time Division Multiplexing is done in digital domain
Digital signals are better to regenerate
Performance monitoring
Ease of encryption (digital can be scrambled easier)
Advantages of Digital
• DSP chips have enabled the transition to digital
networking
– Echo cans now use DSP algorithms (LMS)
– Modems
– Vocoders / Decoders
• Cell phones, Secure phones, Voice over packet, etc.
• There are a few drawbacks to digital
–
–
–
–
Bandwidth management
Network synchronization
Analog interfaces
Multi-access is complicated (drop and insert)
Switched voice architecture
From: Digital Telephony Bellamy, chapter 1
Keshav, chapter2
The voice network
Bandwidth is
allocated
through the
network using a
parallel SS7
network.
Public Switch
Public Switch
Interexchange toll
switch network
Public Switch
Tandem switch
Central office
Telephone
Telephone
Telephone
Central office
From: Voice over IP Fundamentals, Cisco
Press
The SS7 Network
• SSP: Service switching point. Originates
the messages requesting bandwidth through
the network.
• STP: Signal transfer points. Packet
switches for signaling messages.
• SCP: Service control points. Servers that
host routing instructions and enhanced
services.
From: Voice over IP Fundamentals, Cisco
Press
Keshav, Chapter 15
SS7 Protocol Details
• Note that SS7 is it’s own protocol with
layers.
– Much like the TCP/IP stack
• Layers 1-3 are MTP and are switched
through STPs (similar to routers)
– Layer 1 is T1 connections
– Layer 3 routes on point codes (like IP address)
SS7 Protocol Details (cont)
• 3 different stacks on top of MTP
– TCAP/SCCP
• Services requiring database lookup:
– Calling cards, interactive dialing
– TUP: Telephone User Part
• Basic telephone services
– ISUP: ISDN User Part
• Enhanced services like:
– User to User signaling, VPNs, Caller ID, Call Forwarding