ethernet - budapest university of technology and...
TRANSCRIPT
Budapest University of Technology and Economics Department of Telecommunications and Media Informatics
Ethernet
Access Technologies 2
Moldován István
BME-TMIT
Ethernet Forwarding
Physical Topology
Active (Spanning Tree) Topology
VLAN Forwarding Topology
MAC Forwarding Topology
BME-TMIT
3
Ethernet Switches
● Layer 2 forwarding – MAC address based
● Learns MAC addresses
● Store-and-forward operation ● No collision
● High speed backplane
● Many interfaces ● Different interface speeds
● Different media
BME-TMIT
4
Switches
● Standard refers them as bridges
● They divide the broadcast domains
● Types
● Unmanaged
– SOHO use, low level aggregation
– No support for STP nor VLAN
● Managed
– VLAN and STP support
– management interface
● Enterprise switches – L2/L3
BME-TMIT
5
P-to-P mode
● On links between bridges CSMA-CD not needed
● Separate RX/TX paths at phy
● No collision
● Full duplex
● Higher achievable BW
● Can be used
● Between bridges
● Between bridge-PC
● HUB and shared media can not use it
BME-TMIT
6
Ethernet local network design
● Hierachically
HUB
SWITCH, Bridge
Multiservice switch
Router
BME-TMIT
Page 7
Metro Ethernet
Metro Backbone
National IP
Network
GE
Aggregation
Switch
2 x GE
GE
GE
VOD
Server
TV Headend
3rd Party
ISP
VOD
Server VOD
Server
POP
B-RAS
Router
Internet
E.PON EFM G.PON
100BaseFX
B.PON
GE GE
GE
GE
Location Server
SIP Proxy
Server
Triple Play
BME-TMIT
Ethernet VPN services
● E-Line (MEF) [ITU: Ethernet Virtual Private Line EVPL, IETF: Virtual Private Wire Service, VPWS] ● Leased line
● Point-to-point
● E-LAN (MEF) [ITU: Ethernet Virtual Private LAN EVPLAN, IETF: Virtual Private LAN Service, VPLS] ● Virtual LAN service
● Multipoint-to-multipoint
● UNI-connection point ● Virtual connection - EVC
Source: MEF
BME-TMIT
What is a “LAN-like Ethernet service”? Definition of L2 VPN
● L2 VPN: connection of multiple sites in a single bridged domain over a SP network
● Customer perspective:
● all sites appear to be connected to a single Ethernet-Switch/Segment
● no L2 protocol conversion between LAN/WAN
● no knowledge required on WAN technologies (e.g. FR)
● complete control and freedom of routing (IP, IPX, AppleT, DecN, etc.)
● simple to add new sites: no reconfiguration at existing ones
● SP perspective:
● logical separation of the existing network resources in order to provide L2 connectivity
BME-TMIT
.1 .2
.5 .4 .3
P2P versus MP Connectivity
● P2P
● customer edge (CE) end point or node are able to communicate to a single other CE nodes.
● P2MP (a.k.a. hub-and-spoke)
● end customer designates one CE node to be the hub that multiplexes multiple point-to-point services over a single User-Network Interface (UNI) to reach multiple "spoke" CE nodes. This means that each spoke can reach any other spoke only by communicating through the hub.
● MP
● CE end point or node is able to communicate directly and independently to all other CE nodes.
● Three generations of Ethernet
● FE: CSMA/CD*
● GE: CSMA/CD + Full-Duplex
● 10GE: Full-duplex Only
● MP by nature
● CSMA/CD:
– Carrier Sense Multiple Access with Collision Detection
● Full-Duplex (on P2P links)
– L2 and L3 communication = MP
● L2/L3 protocols
● assumes MP connectivity by default on Ethernet interfaces
BME-TMIT
Router Inter-connect Characteristics
● Most prevalent L2 VPN service
● Connect Routers
● Routers provide clear demarcation
– Possibility of unnecessary BC/MC eliminated
– L2CP issues eliminated
– One MAC address per site
● Most customers use routers to connect multiple sites together
● Well-known design
● Considerably reduces management overhead (single connection per site)
BME-TMIT
● L2 VPN service should be agnostic to customer’s L3 traffic (e.g. IPv4, IPv6, IPX, AppleTalk, etc.) encapsulated within L2 frames
Router Inter-connect Requirements
● Traffic type:
● Unicast, multicast, broadcast
● Efficient replication of MC/BC traffic is desirable
● Security
● Traffic separation for customer’s of different L2 VPNs
● Similar characteristics as for e.g FR
● Agnostic to security mechanisms employed at L3 and higher layers
● Addressing
● Independent addressing (overlapping) should be supported
– L2: MAC, VLAN ID;
– L3: IP
● VLAN ID preservation/translation
● QoS
● Best effort, delay sensitive, loss sensitive traffic
● Queuing and forwarding policies should preserve packet order for packets with same QoS parameters
● Protection and restoration
● Redundant paths to assure high availability
● Reaction to failure: restore the service using alternative paths
● Restoration time < L3RP detection time
BME-TMIT
Switch Inter-connect Characteristics
● Connect L2 switches (L2CP issues arise)
● Multiple MAC address per site
● Potential scalability issues
● Malfunctioning L2 switches can also flood the provider network with BC/MC traffic
● Requires controls of FIB size (# MAC per site) and BC/MC rate-limiting
● Should be sold as a premium service
● Charge customers per site per block of MAC addresses
BME-TMIT
● L2 VPN service should be agnostic to customer’s L3 traffic (e.g. IPv4, IPv6, IPX, AppleTalk, etc.) encapsulated within L2 frames
● L2 VPN service should not interfere with existing L2 protocols
Switch Inter-connect Requirements
● Traffic type: ● Unicast, multicast, broadcast ● Efficient replication of MC/BC traffic
is desirable
● Security ● Traffic separation for customer’s of
different L2 VPNs (like e.g FR) ● Agnostic to security mechanisms
employed at L3 and higher layers ● L2 security mechanisms (e.g.
802.10b) may inhibit L2 VPN service, when VLAN IDs are encrypted
● Addressing ● Independent addressing
(overlapping) should be supported – L2: MAC, VLAN ID;
● VLAN ID preservation/translation – Translation affects e.g. MST
(802.1s)
● QoS ● Best effort, delay sensitive, loss
sensitive traffic ● Queuing and forwarding policies
should preserve packet order for packets with same QoS parameters
● Protection and restoration ● Redundant paths to assure high
availability ● Reaction to failure: restore the
service using alternative paths ● Restoration time < CE_device,
L2CP, L3RP detection time
● L2CP ● Support for transparent operation of
L2CPs employed by the customer
Budapest University of Technology and Economics Department of Telecommunications and Media Informatics
Ethernet based transport in provider networks
The Ethernet way
BME-TMIT
Challenges
● Carrier grade requirements
● Scalability
● Service Quality
● Multicast
● Management
– Fault
– performance
Upgrade the cheap Ethernet (IEEE)
Extend the proven MPLS (IETF)
Provide Ethernet service
BME-TMIT
Carrier Ethernet: service types
●E-Line service:
● Ethernet Private Line
● Virtual Private Line
● Ethernet Internet Access
●E-LAN Service:
● Multipoint L2 VPN
● Transzparent LAN
● Needed for IPTV multicast etc
E-Line
E-LAN
Point-to-Point EVC
Carrier Ethernet Network
UNI: User Network Interface, CE: Customer
Equipment
CE
UNI UNI
CE
Multipoint-to-Multipoint EVC
Carrier Ethernet Network
CE
UNI
MEF által hitelesített Carrier Ethernet
termékek
CE
UNI
BME-TMIT
IEEE 802.1Q - VLAN
● VLAN tag
● QoS: priority
● 12 bit VLAN ID: 4096 VLANs
● Usage
● User identification
● Service identification
● The 4096 limit is there– Too few for a provider!
● The most wide spread UNI
● Also we must be prepared to transfer VLAN tagged packets
MAC
DA
MAC
SA
802.1Q
ethertype VLAN tag Data CRC
VLAN ID
(12 bit)
Pri
(3 bits)
CFI
(1 bit)
BME-TMIT
Provider Bridges (IEEE 802.1ad)
● Also known as Q-in-Q
● Widely used
● 4K services (12-bits)
● Unique service ID
● (S-VID) ● Forwarding is the same, L2 learning bridge with STP, filtering for the outer
VLAN (S-VID)
● Scalability
● 4K service
BME-TMIT
Provider Backbone Bridges
● 4K connected LAN
● Unique per service ID
● (LAN = I-SID)
● Forwarding is the same, L2 learning bridge with STP, filtering for the outer VLAN (B-VID)
● Service management is simple
● Scalability
● Massive sservice
scalability (24-bit)
● Only learn MAC of the Provider bridges
● Mapping of C-MAC to VIDs
BME-TMIT
Comparison – headers added
BME-TMIT
PB/PBB facts
● Scalability solved
● Cheap Ethernet switching remains
● Still no support for Traffic Engineering
● Protection/restoration based on STP
● Management is more complex ● Different layers of VLANs
● No adequate management
● Still not good in the core…
BME-TMIT
PBT ● Goal
● Keep the Ethernet forwarding
● Change the control plane (no STP and learning)
● Set up paths ”manually“
= Traffic Engineering - Ethernet
● What we get:
● Point-point tunnel
● Traffic Engineering
● Protection
BVID=1
BVID=2
BVID=1
BVID=2
BME-TMIT
PBT ● Provider Backbone Transport – IEEE 802.1Qay
● Nortel started
● based on PBB
● Uses the existing technologies
● Deterministic QoS for service is the target
● scalability
BME-TMIT
PBT - operation ● Data plane
● Static forwarding tables
● Addressing
– 60 bit MAC + VLAN
based
● Totally different control plane
● Manual
● MPLS based
BME-TMIT
Ethernet Transport technology use
IP/MPLS mag
Voice
Data
Video PB
PBB
PBT 802.1Q
802.1Q: 4K user
PB: 4K service, Not too many MAC
Added value: TE, OAM
PBB: good scalability
PB – Q-in-Q – IEEE 802.1ad PBB – Mac-in-Mac – IEEE 802.1ah PBT – PBB-TE – IEEE 802.1Qay
Budapest University of Technology and Economics Department of Telecommunications and Media Informatics
Ethernet based transport in provider networks
The other way – IP/MPLS
BME-TMIT
28 | Áramkör kapcsolt Ethernet? | Moldován István | Elsinco szeminárium | 12 Szeptember 2007
MPLS Pseudowire - WPWS ● Ethernet p2p service
● IETF pwe3 study group, the draft name Martini –encapsulation
● MPLS label is encapsulated, multiple virtual connections within an UNI (VC)
● Forwarding based on tunnel label
● The solution inherits all MPLS solutions
● Traffic Engineering, protection, OAM
CE CEPE PESP
CE CEPE PESP
BME-TMIT
Ethernet Szolgáltatások 29 [email protected]
Reserved Sequence Number
EXP TTL 1 Label (VC)
EXP TTL 0 Label (Tunnel)
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
Layer2 PDU
Control
Word
VC Label
L2 header
Original Ethernet Frame Tunnel Label
Data Plane : EoMPLS packet
● Tunnel label :
● LSP label to get the packet from ingress PE to egress PE (IGP label or RSVP (TE) label)
● VC Label :
● demultiplexing label identifying an emulated VC
● Identifies outgoing interface/vlan
● Control Word : extra information regarding the VC
● VC Label TTL = 2
BME-TMIT
Ethernet Szolgáltatások 30 [email protected]
Pseudo-Wire reference model |<--- Emulated Service:FR/Ether/ATM/PPP/HDLC --->|
|<------ PW:Pseudowire VC ----->| |<-- FR -->|
Ethernet
ATM/PPP/HDLC
|<--FR -->|
Ethernet
ATM/PPP/HDLC
ES Emulated Services: FR/Ether/ATM/PPP/HDLC
Attachment VC (AVC): FR DLCI/Ethernet VLAN/ATM PVC/PPP/HDLC
PW Pseudo-Wire: Emulated VC (EVC): MPLS LSP
PSN Packet Switched Network (Tunnel): MPLS LSP or RSVP-TE
IP/MPLS
Core
Attachment VC /
L2 circuit
Attachment VC /
L2 circuit
Attachment VC /
L2 circuit
Site1A
Site 2A
Attachment VC /
L2 circuit PE1 Site1B
Site 2B
PE2 CE 1A
CE 2A
CE 1B
CE 2B
PSN Tunnel: (IP/)MPLS Tunnel
MPLS (LDP or RSVP-TE)
BME-TMIT
Ethernet Szolgáltatások 31 [email protected]
MTU Calculations for EoMPLS:
Max Frame Size = Link Header + labels + Transported L2 Header + Payload
Transported Ethernet Header:
AToM removes (1) Preamble (2) SFD (3) FCS
Ethernet II Encapsulation 18 Bytes
Ethernet SNAP 26 Bytes
Dot1q tag(s) 4 Bytes per tag
Labels : usually 2 labels
Example : Ethernet II + dot1q tag + 2 labels + Ethernet II + 2 dot1q tags (QinQ) + Payload 18B 4B 8B 18B 8B 1500B
Comment on VPWS: MTU
● EoMPLS does not support fragmentation
● MTU > layer2 VLAN frame
● No e2e detection
● MTU in core should be bigger
● MTU values
● a PE-CE should match
● MTU set them correctly
BME-TMIT
Ethernet Szolgáltatások 32 [email protected]
VPLS
PE
Service Provider Backbone
PE
PE
VPLS - A
VPLS - B
VPLS - B
VPLS - A Emulated LAN
CE - 1 CE - 2’
CE - 2
CE - 1’
Bridged LAN
Customer Edges (CE): Client side device, tyically Ethernet Provider Edges (PE): VPLS inteligence, start/end Core: just forwarding
BME-TMIT
Ethernet Szolgáltatások 33 [email protected]
VPLS example
Full Mesh PEs are acting like a bridge towards the CE nodes
BME-TMIT
Ethernet Szolgáltatások 34 [email protected]
VPLS Operation
● VPLS instance : Service–identifier (Svc-id)
● Full mesh tunnels
● Targeted LDP messages
● Forwarding: learning bridge
● Flooding
● Split-horizon – never send to the receiving interface
BME-TMIT
Ethernet Szolgáltatások 35 [email protected]
Number of
PE in
VLAN
Number of
LSPs
Number of
Retransmissions/
Broadcast
5 20 4
10 90 9
20 380 19
40 1 560 39
Why not VPLS End-to-End?
● VPLS scalability
● eg. 5 PE - 20 LSP, 40 PE: 1,536.
● High bandwidth waste because of broadcasts
● VPLS – new requirements
● Protection, OAM, mapping
BME-TMIT
Ethernet Szolgáltatások 36 [email protected]
MPLS core
PE
CE CE
PE
CE
MPLS core
PE
CE CE
PE
CE
E.g ARP-request
E.g ARP-reply
VPLS – Flooding & forwarding
● Flooding (Broadcast, Multicast, Unknown Unicast)
● Dynamic learning of MAC addresses on PHY and VCs
● Forwarding
– Physical Port
– Virtual Circuit
BME-TMIT
Ethernet Szolgáltatások 37 [email protected]
VPLS scalability- hierarchical
● MTU - Multi-Tenant Unit: owned by multiple users, bridge
● VPLS can be extended to the MTUs
● MAC/VLAN scalability increased
● More complex MTU
● Hierarchical VPLS
● „HUB” pseudowire (hub PW) between PEs
● „spoke” PW between MTU-PE
– Spoke PW can be QiQ, MPLS, …
BME-TMIT
Ethernet Szolgáltatások 39 [email protected]
H-VPLS
MPLS core MPLS edge Ethernet edge
p2p or ring
H-VPLS
n-PE
u-PE u-PE
n-PE
MPLS core MPLS edge Ethernet edge
VPLS
n-PE
CE CE
n-PE
VPLS – Architectures
● VPLS
● One big hierarchy
● MPLS to the Edge
● H-VPLS
● 2 level Hierarchy
● MPLS or Ethernet Edge
● MPLS core
BME-TMIT
Ethernet Szolgáltatások 40 [email protected]
VPLS proposal Auto-discovery Signalling / label distribution
Draft Kompella VPLS BGP BGP
Draft Lasserre-Vkompella
VPLS
None (several options
possible)
LDP
VPN Discovery
Signaling
Centralized DNS, LDAP, Radius Directory Services
Distributed BGP
Label Distribution Protocol (LDP, BGP)
VPLS signaling and auto-discovery
● VPLS requires full mesh of LSPs between PEs:
● Manual procedures (static)
● Provisioning systems(NMS/OSS)
● Signalling protocols:
– LDP (“Lasserre-V. Kompella” draft)
– BGP ( “Kompella” draft, Juniper)
– other (Radius, DNS, stb.)
BME-TMIT
Ethernet Szolgáltatások 41 [email protected]
WDM
SDH
RPR
Ethernet
Over
MPLS
Q-in-Q
VPLS
VPWS
L2TPv3
L2VPN
Customers do NOT
care about technology!
They are interested
in the SERVICE!
Ethernet transport alternatives- Ethernet over anything
● Ethernet over legacy networks
● ATM: rfc2684-B
● FR: rfc2427-B
● PPP: rfc2878
● Ethernet over Ethernet
● QinQ, MACinMAC
● Ethernet over SDH
● GFP, VCAT, LCAS
● Ethernet over IP/MPLS
● L2TPv3
● VPWS, VPLS
● Ethernet over WDM
● Ethernet over RPR
● Which technology to use?
● What Service to offer?
● Depend:
● SP strategy
● Service definition
● Existing investments
BME-TMIT
Conclusions
● All are assuming Ethernet aggregation
● All provide basic Ethernet level connectivity
● All provide similar ways for service identfication
● S-VLANs
● PPPoE still can be used
● Support for Multicast
● QoS handling is similar
– Or can be mapped (DSL line PVCs-> VLANs)
● No specific requirements!
42
Budapest University of Technology and Economics Department of Telecommunications and Media Informatics
Thank You for your attention