fixed network infrastructure - nus computing
TRANSCRIPT
Ivan Tam 2015
Fixed Network Infrastructure – An OverviewIVAN TAM
Applications, Broadband, Fixed Mobile Infrastructures
Web browsing, on-line service, peer to peer, video streaming, 4K, Cloud Storage,
Wired Access – Most cost effective technology over wire to connect home/enterprise over high speed
IP transport – Transport routing from service provider network to core/data center and Internet
Cloud, Data Center
Wavelength Division Multiplexing
Wireless Access – Most cost effective technology over limited spectrum to connect user wirelessly and maintaining mobility
Internet peering and transit - A number of carriers to the final network of the content
Caching, local content, Billing and charging
Tracking user, keeping record of service profile, authentication
Bandwidth Management
Agenda
Access Technologies Copper based xDSL, Fiber based PON technology
Optical Networking Technology
IP Networks IP Routing in carrier
Broadband Network Gateway (BNG) for Residential services
QoS
Core and Internet Gateway
Service Provider Architecture
InternetCoreService EdgeAggregationAccess
xDSL
Core RouterInternet Gateway
Transit Carrier Router
Internet Exchanges
• Access – provides cost/performance effective connectivity using copper, fiber to home and offices
• Aggregation –aggregate the traffic from various types of access terminating at CO/POP
• Service/edge –service management point where authentication, service quality, and service features are provided BNG ( Broadband Gateway)/VPN PE (Provider Edge)
• AAA – where database of customers are stored and checked by service edge
• Value added service (VAS) –IPTV server, internet cache
• Core – national connectivity between data centers and service edge to each other and to the internet
• Transit – service provider interface with a number of transit carriers via own border routers. Transit carrier is responsible for routing traffic back and forth to destinations in the worls
Data Center
Central Office
Transit
Web site’s service provider
Web site’s hosting data center/ Cloud
Physical Infrastructure
InternetCoreService EdgeAggregation/Metro IPAccess
GPON
xDSL
Metro-ethernet
Core Router
Internet Gateway
BroadbandGateway/VPN Provider Edge
Access Node
Central OfficeData Center
AAA/VAS
Transit Carrier Router
Internet ExchangesSource: ECMweb.com
• CO (Central Office) -distributed throughout a city terminating copper pairs from homes and offices within a few km
• PSTN switches (aka telephone switch) -telephony service, ADSL DSLAM -broadband service
• 10s of COs in a typical city• FTTx (fiber to cabinet,
building, home) => termination of copper pair moves out of CO and fiber from them enables CO consolidation
• CO are connected to Metro Data center using metro IP transport network
Access Technologies Leverage on existing infrastructure such as telephone twisted pair (CAT 3) or cable TV coaxial cable HFC (Hybrid Fiber Coaxial network) to deliver very high speed access Reduce or delay investment to lay new wiring (e.g., fiber)
Or simply deploy fiber all the way to home outright (FTTH) Gbps, and 10Gbps of bandwidth to homes
Shared for economics, and easy to maintain
PON (passive optical network)
Evolution of DSP enables more advanced techniques to be used Use more spectrum of the copper wire, xDSL (Digital Subscriber Loop)
Divide the spectrum into channels
Invariably due to attenuation and use of high frequencies in the spectrum Trade off of distance and bandwidth -> reduce distance of copper/coaxial
Termination equipment is closer to home, fiber need to extend further from network (FTTX)
xDSL Technologies (1) Broadband over existing telephone copper pair
E.g., Cat 3
Evolve starting from ~1997 ADSL (Asymmetric) ADSL2, ADSL2+ (u=24, d=3.3 Mbps), VDSL2 (u=100,
d=100/50Mbps), vectoring and G.Fast
VDSL2 profiles support different speed and deployment scenario
Wider and wider spectrum gradually used
DMT (discrete Multi-tone) Divides spectrum into channels (4.12135kHz)
Allow modulation scheme to be adapted according to noise situation at different part of spectrum
Quality and speed depends on: Distance sensitive due to attenuation
Copper wire quality, wire gauge, un-sealed termination, bridge taps
Cross-talk is one of the major issues
Near end cross talk (Next), Far end cross talk (FEXT)
Robustness improvement techniques – Interleaving, FEC
Binder25xTwisted Pair
Cable
DSLAM
ModemFEXTNEXT
Central Office /Cabinet Home
Home
Home
4KHz 0.14Mhz 1.1Mhz 2.2Mhz 8.8Mhz 12Mhz 17Mz 30Mhz
ADSL2+
VDSL2 frequency Profiles
ADSL2UpstreamPOTS
VDSL2 17A VDSL2 30A
0.276Mhz (annex M) VDSL – divides spectrum into multi-band for upstream, downstreamchannel width at 8K for Profile 30a
VDSL2 12AVDSL2 8A
Earlier xDSL Technologies
25Mbps
50Mbps
75Mbps
100Mbps
10 Mbps
1 km 2 km 3 km 4 km 5 km
24 Mbps
16 Mbps
8 Mbps
VDSL2 (performance subjected to FEXT)
ADSL2
ADSL
12.5Mbps
FTTC (Fiber to Curb)
FTTB (Fiber to Building)
ADSL2+
Copper connection
xDSL with Fiber uplink
powerCentral Office
xDSL Technologies (2) Vectoring – cancelling of x-talk effects
Theoretical speed of VDSL2 often not achievable due to cross talk
Computation of x-talk effect at far end and pre-compensate them, like noise cancellation
Push up the downstream speed of VDSL-2 to true ~ 150Mbps up to 400meter
G.fast – use spectrum up to 106Mhz Achieve 500-1000Mbps <100m, ~200 Mbps @ 200m
Vectoring is a must due to use of high frequency
ADSL
ADSL2+VDSL2 8b
VDSL2 17a
VDSL2
30 MHz
G.fast
106 MHz
G.fast
212 MHz
0 50 100 150 200
bandwidth [MHz]
ADSL
ADSL2+
VDSL2 8b
VDSL2 17a
VDSL2 30a
G.fast 100 MHz
G.fast 200 MHz
Modem
FEXT
FEXT from line 1 effect on Line-1 is considered, and certain pattern is preimposed
The x-talk effect from line 1 is cancelled
Line 1
Line 2
Time division (TDD) based for up and down steam
Unlike previous xDSL which is based on frequency band (FDD)
Fiber further deeper and near to home -> FTTdp (Fiber to the distribution point), ease of deployment form factor
212 Mhz coming
400-700 meters
FTTdp
Modem< 200 meters
Ease of Deployment
PON Technologies (1) PON (Passive Optical Network) ePON and GPON
GPON (Gigabit Passive Optical Networks) ITU G.984.1 – 984.6
Over single core fiber shared up to the splitting point Splitting ratio 1:32, 1:64, 1:128
Optical loss over distance, and over splitter
OLT (optical line termination) at the network side and ONT(optical network termination) at home
GPON protocol defined how the shared fiber medium is accessed Broadcast to all in downstream, but frames are identified for
individual ONT, upstream is based on TDMA (TDM Access)
DBA (dynamic bandwidth allocation) to control bandwidth and slots allocated to upstream, re-allocate every few ms
QOS by defining guaranteed, available, and best effort allocation mechanism using 5 types of T-CONT implemented by time slot allocation
Each ONT can have one or more T-CONT and identified on per ONT basis
OLT
2.5Gbps Shared
1.25 Gbps Shared
20 km
ONTFeeder Fiber (single core G.652)
1310 nm
1490 nm
Splitter
Distribution fiber
Drop fiber
GTC PayloadPCBd GTC PayloadPCBd GTC PayloadPCBd
Port_id Payload
GEM Frame
Port_id PayloadSync/OAM/Error Check US BwMap
Upstream Bandwidth Map
Alloc_Id Start Stop
125 us
Header Payload Header Payload Header Payload
Sync, OAM ONT T-CONT Buffer fill level
38880 bits
125 us19440 bits
Up stream
Down stream
Frame Structure (Simplified)
PON (Passive Optical Network) Deployment
OLT
Downstream bandwidth 2.5Gbps Shared
Upstream bandwidth = 1.25 Gbps Shared
< 20 km
ONT
Feeder Fiber (single core G.652)
Drop fiber1310 nm
1490 nm
Splitter
Distribution fiber
Fiber loss/km~0.36db
1:2 splitter~3.2db
1:32 splitter~16.5 db
Connector~0.4 db
Class B: 28.5db, Class C+: 32.5db
PON Technologies (2) NG PON1 – 10Gbps down 2.5Gbps up, coexists with current fiber and GPON, under ITU G.987.x Uses current fiber and splitter, coexists with GPON
Very important in keeping current investment
1:256 split, 40km
NG PON2 - TWDM (Time Division and Wavelength Multiplexed) PON under ITU G.989.x 4 or more wavelength in a fiber each λ runs 10Gbps
symmetrical
An ONT is allocated a single wavelength but share with other ONTs allocated the same wavelength
Start with a single wavelength and incrementally add in additional as needed
Cost effective Wavelength tunable ONT is needed
Alternative is to use fixed lambda
10Gbps applications Mostly used to backhaul other fttx equipment back to
network
Mobile backhaul for LTE-Advanced and beyond
OLT
10 Gbps Shared per λ
10 Gbps Shared per λ
40 km
ONT
1300 - 1320 1480-1500 1530 - 1540 1595 - 1625
Splitter
1260-1280 1575-1580
XGPON1 up GPON up GPON Dn NGPON2 up xGPON1 Dn NGPON2 DN
Co-exist and Mux and Demx of GPON, XG-PON1, NGPON2 λ to different line cards and
GPON, XG-PON1, NG-PON2 coexists
4 λ of NG-PON2
L Band1565 1625
C Band15301460
S Band1360
E BandO Band1260
Optical Networking Technologies
Wavelength Division Multiplexing Cost of deploying fiber is high, and takes time, e.g., in urban areas Economize on use of fiber, one pair for each router port connection is
expensive Transmit information on different wavelengths What determine how much we can transmit per wavelength, can we share
them? Define what wavelength are to be used, how much of it, and standardize
them
Optical Network Point to point fiber technology is simple but building fiber network this way is
expensive Ring provides redundancy, allow longer distance fiber to cover more stops Hence the function of add-drop is needed to share the pair of fiber But if more redundancy and larger area is needed, mesh topology is more
suitable Wavelength routing technology is needed to route a wavelength across a
mesh topology
Two Fiber pairs
DWDM
Two wavelength over one fiber pair
Point to point
Ring
Mesh
WDM Technologies (1) Carrying multiple router connections using different wavelengths
e.g., up to 8.8Tbps per fiber in case of 88 channels of 100Gbps each, or even 17.6 Tbps per fiber using 88 channels of 200Gbps
Normal 10Gbe optics transmits at 1310nm (LR -10km) or 1550nm (ER, ZR 40-80Km) over 2 core
DWDM (Dense Wavelength Division Multiplexing) Based on ITU G.694.1 grid of 12.5, 25, 50, 100Ghz channels plan at
C-Band and L-band
Up to 96 channels (currently) per fiber
High grade quality optics due to narrower channel width
Used where fiber is scarce, e.g. metro, long haul, and submarine cable
CWDM (Coarse Wavelength Division Multiplexing) Based on ITU G.694.2 grid of 20nm channel plan
Up to 16 channels, mostly 8 deployed
Relaxed channel width, less precise and cheaper optics can be used
Deployed where distance is short, economize on fiber usage is good but not at high cost, e.g., access
OTN (Optical Transport Network) Standard ◦ Used for multiplexing lower speed payload to higher speed
channel to more effectively utilize high speed λ
◦ ITU-G.709, cf. http://www.itu.int/ITU-T/studygroups/com15/otn/OTNtutorial.pdf
Mux/Demux(4, 8, 40 λ)Tx
Rcv
ProcessingMappingFEC
Tx
Rcv
ProcessingMapping
Rcv
Tx
ProcessingMapping
Rcv
Tx
ProcessingMapping
R1
R2
R3
R4
R5
R1
R2
R3
R4
R5
Amplifier for long distance
Normal B&W opticse.g., 10Gbe
Converting multiplexed OTU to Och (colored optics)e.g., 100 G λ
IP/Ethernet
10Gbe -> ODU-2
10Gbe -> ODU-2
ODU-4 OTU-4
OTU = Management overhead + ODU + FECe.g., OUT-4 > 100Gbps due to above
ODU-0=1Gbps, ODU1=2.5Gbps,ODU-2=10Gbps, ODU-3=40Gbps, ODU4=100Gbps
OchOTN
DWDM Transponder Transponder
Wavelength Division Multiplexing (C-band)
L Band1565 1625
C Band15301460
S Band1360
E BandO Band1260
19
1.3
5TH
z 1
56
6.7
2nm
19
1.4
THz
1
56
6.3
1nm
19
3.1
THz
1
55
2.5
2nm
19
3.1
5TH
z 1
55
2.1
2nm
19
3.0
5TH
z 1
55
2.9
3nm
19
6.0
5TH
z 1
52
9.1
6nm
19
6.1
THz
1
52
8.7
7nm
19
3.1
THz
1
55
2.5
2nm
19
3.2
THz
1
55
1.7
2nm
19
3.2
THz
15
55
1.7
2nm
19
5.9
THz
1
53
0.3
3nm
19
2TH
z
15
36
1.4
2n
m
193.1THz anchor
12
71
nm
12
91
nm
15
91
nm
16
11
nm
14
51
nm
14
71
nm
15
51
nm
15
71
nm
19
1.8
THz
15
36
3.0
5nm
19
6.1
THz
15
28
.77
nm
44 Channels40 channels
96 channels
16 channels
10 Gbe, 40 Gbe
10 Gbe, 40 Gbe, 100Gbe
CWDM
DWDM50Ghz
DWDM100Ghz
ITU G.694.1, 694.2 Grid
+6/7nmEach side, rest is for separation
Ring Topology and Protection
TransponderOADM
TransponderOADM
R1 R2
R3 R4
R5 R6R1 R2
R3 R4
R5 R6
Protected Unprotected(rely on dual routers)
Ring topology offers optics and fiber protection when traffic is sent via east and west direction
This can be done via using Y cable where incoming traffic is splited into two
transponders each transmit and receive in different direction
At the receiving side, only one of the incoming signal is selected and passed out of the node
Alternatively, one can simply use two routers, or a single router with two ports
Y cable show is actually deployed as a pair, one for splitting signal from input and the other for
Y-cable
Optical Network transport
InternetCoreService EdgeAggregationAccess
• Metro-transport • Transport and
aggregate the traffic from access to the core/data center
• Expected to grow very significantly as the access bandwidth increases and traffic consumption via video streaming increases
• Most cost effective ways of moving bits between access and data center
Core Router
Internet Gateway
BroadbandGateway/VPN Provider Edge
Core/Data Center
AAA/VAS
Transit Carrier Router
Internet Exchanges
GPON
Metro-ethernet
Central Office
xDSL
Access Node
• Fiber Path available ? Count increases
• Distance too far
• Port Count increases as traffic grows• One fiber per 10Gbe port?• Group them into 100Gbps and transmit
Optical Network transport
InternetCoreService EdgeAggregationAccess
• Metro DWDM transport • The example shown
here consists of two Metro-DWDM rings, each with two central offices and one core nodes
• A ring provides two paths connecting a CO to the core
• A core DWDM ring is also show connecting the two data centers and the internet exchanges
Core Router
Internet Gateway
Core/Data Center
AAA/VAS
Transit Carrier Router
Internet Exchanges
GPON
Metro-ethernet
Central Office
xDSL
Access Node
Core DWDM Ring
Metro DWDM Ring
WDM Technologies (2) Coherent Transmission
Major advancement for 100Gbps and beyond
Polarization, high order modulation, coherent detection, better DSP
ROADM (Reconfigurable OADM) Colorless – any λ can be add-drop
Directionless – traffic can go any direction
Contentionless – any λ (overlapping) can be add-drop at different direction
No manual work for above!
Based on WSS (Wavelength Selective Switch)
Enable True wavelength routing, dynamic set up, recovery in mesh network, improving redundancy and wavelength efficiency
FlexiGrid ITU defined granularity at 12.5Ghz, nx12.5GHz to form
“superchannel”, allowing more flexible allocations and improve efficiency
E.g. possible to carry 100Gbps over a 37.5Ghz superchannel, 400 Gbps is carried as 100Ghz super channel rather than 4 x50Ghz 100Gbps channel
Because the of the arbitrary width of superchannel, flexgridrequires new filter, wss and transceiver
50Ghz 100Gbps
100Gbps /37.5Ghz 400Gbps /100Ghz 1Tbps /200Ghz
WSS
4-Degree WSS
R1 R2
R3
R4
R5
R1
R2
R3
R4
R5
Site A
Site B
Site C
Site D
Site E
Site F
Internet Routing (conceptual)IP Routing among service providers (EGP)
Border routers of service provider are interconnected
Border router shares the set of IP addresses that it’s own service provider can reach
From connected foreign border routers, learn what are the IP addresses other service providers can reach
Forward Internet packet to the selected foreign border router
Not aware of how it is routed subsequently, no visibility to the topologies of other service providers
Service provider fully control the routing within is own domain called AS (Autonomous System), with unique AS number
E.g., BGP
IP Routing within a Service Provider (IGP) By having router to share IP addresses and its link costs among routers within the
same service provider
The IP addresses (network part) and network topology is visible to all
A router determine the “shortest” path to route a given IP address
IGP Protocol, E.g., OSPF, IS-IS
SP1
SP2
SP3
SP4
IP addresses that I can reach, include those from mine SP2 , and SP3
Web site’s hosting data center/ Cloud
Management Port
Control Card
Line Cards
Fan Tray
Switching Cards
Line Card – Ingress • De-framing• Classification of
traffic by QoS• Look up on
destination identify port/card for egress
• Execute any configured policy
• Buffering and transmit to the switching card
• Maintain forward table in conjunction with control card
• Collect statistics
Line Cards
Line Card - Egress• Accept packet or frame from
switching card• Queuing and buffering according to
class of service at the egress port• Framing for transmission• Collect statistics
• Run routing protocol and other service protocol
• Formulate routing table and forwarding table
• Accept management command for configuration
• Monitor the status of the equipment
• Switch packet or frame to the destination line card
Anatomy Of A Router
L2 vs. L3 L2 based on MAC addresses (48 bits) A switch keeps track of which direction to forward for MAC addresses
Broadcast to all directions when it does not know how to forward
Spanning tree is used to restrict broadcast and avoid looping and storm
Learn the direction when the destined terminal sends return frame
VLAN provides “domain” separation, L2 traffic broadcasted within VLAN, often correspond to a IP subnet (see below)
L3 based on IP addresses 32 bit of IP addresses has network and host part, routing requires knowing
where to direct based on network address (IP subnet), same for IPv6 (128bit)
IP scalability allows us to build Internet, only network part of IP address is learned (best practice)
In IGP - A router keeps routing table, exchange IP addresses and topologies via routing Protocol
When link failure is noted, router sends via other path to destination
IP Addresses can be public which is unique worldwide and used in Internet assigned by Internet Authorities to service providers and then to customers
IP Address can also be private overlapping, e.g., 10.0.0.0/8, 192.168.0.0/16, and reused by different organizations internally
B
1 2
1 2 1 2
Routing Protocol
VLAN X VLAN Y
R1 R2
R3
MAC Addresses
L3 Routing Table
IP Network Part
A
C
1
L2 Forwarding Table
A
C
1
Subnet
InternetCoreService EdgeAggregationAccess
• iBGP – shared the learned IP addresses from border gateway within the operator’s own router network, allow internet packet to be routed
xDSLIGP, e.g., OSPF/ISIS
EGP, i.e., BGPiBGP
L2
• IGP – routing of IP addresses within the operator domain, e.g., OSPF, IS-IS
• Each router has full visibility of links, routers, and what addresses are reachable behind router
• For each IP network addresses that it is aware of from the IGP, each router find the shortest path from itself to the destination router
• Any packet coming in with that IP address is routed out to the first link on the path
• BGP – allows an operator to tell another operator what IP networks addresses (Prefix) it can reach
• Exchanged between border routers of different AS (Autonomous System)
• To go to IP destination belonging to other operators, a service provider route the packet to the “right” border router, based on IP addresses it learned via BGP
• The border router takes it from there
• Access and Aggregation –normally switch packet by L2, either up or down. Simple configuration
Routing Architecture (example)
Version (4)
Header Length (4)
Priority and Type of Service (8) Total Length (16)
Identification(16)Flat (3)
Fragment Offset (13)
Time to live (8) Protocol (8) Header Checksum (16)
Source IP Address (32)
Destination IP Address (32)
Data
IP Option (0 or 32)
Bit 0 Bit 15 Bit 31
IPv4 Header
QOS Architecture (example)
InternetCoreService EdgeAggregationAccess
QOS Classes• QOS are supported at Ethernet
level (L2) by p-bits (3) in the header, also supported at IP level by DSCP bits (6) in the header
• Up to 8 classes of QoS are usually defined in deployment
• Highest two class are devoted to service provider’s own control and management traffic
• Voice is usually given highest class out of the remaining due low latency requirement
• Video such as IPTV is usually given the second highest, to avoid high lost impacting visual experience
• Important data, e.g., corporate data could also be given a separate class
• Finally, normal internet traffic (including OTT) is given BE (Best effort service)
xDSL L2
ONT remark upstream class of traffic DSCP and 802.1 p-bits when relaying incoming traffic from home. Work with OLT to schedule upstream transmission
OLT schedule upstream with ONT via DBA, manage downstream traffic on a per ONT or even per port per ONT basis based on profile and classes of traffic
BNG keeps subscriber profiles which defines the policy for up and down stream bandwidth or total data. May perform further remarking by looking into the applications type
AAA, Policy
Aggregation router follows the class of service marking and schedule the packet accordingly
Core and border router schedule the traffic according to class of service. If outgoing to overseas, the traffic is passed to transit router.
L3
References and Some Further Info:Fiber Deployment
- https://www.youtube.com/watch?v=a8bzZajwR50
Internet Statistics
- https://www.sandvine.com/trends/global-internet-phenomena/
- https://content.akamai.com/PG2061-SOTI.html?gclid=CMCc3ZCptscCFQ0rjgodTHECSA
- http://www.cisco.com/c/en/us/solutions/service-provider/visual-networking-index-vni/index.html
Internet Exchange example and International Connectivity:
- https://ams-ix.net/
- https://www.telegeography.com/
Network Vision example:
- https://techzine.alcatel-lucent.com/
- http://newsroom.cisco.com/focus
Industry and Standard Organizations
- https://www.broadband-forum.org/
- http://www.ietf.org/
- http://www.itu.int/en/ITU-T/Pages/default.aspx