Transcript Slide 1

Internetworking and QoS
Classical IP
Internetworking and applications QoS
• ATM’s first use was: high-speed backbone for data traffic
– Architectures which allow existing data networks, internetworking
devices and end systems, to connect to an ATM network
• Data architectures
– Create a Layer 3 overlay across the ATM network
– Integrate the ATM layer with Layer 3
• Classical IP over ATM
– ATM as a link-layer
– Gains: support of greater bandwidth, QoS for a connection
• Bridged environment over the ATM (LAN Emulation)
– Provides the link-layer connectivity for Multi-Protocol over ATM (MPOA)
• Peer Model
– Peer the internetwork address with ATM addressing
– Traditional routing protocols control the ATM forwarding path
Internetworking and applications QoS
• LightWeight subnet – minimize the amount of IP header overhead
transported across the ATM network once a binding has been
established via signaling mechanisms
– TCP and UDP aver Lightweight IP (TULIP)
• AAL5 will ensure that packets are not fragmented or misordered, and will
also indicate the packet size
– TCP and UDP over a Non-existed IP Connection (TUNIC)
• No layer 3 information is carried across the ATM network
• Applications will communicate via a dedicated VC utilizing TCP or UDP
directly over AAL5
Internetworking
User
Applications
User
Applications
IP(b) via ATM(b)
IP, IPX, etc
IP, IPX, etc
Layer 3 Routing
OSPF,
NLSP, etc
ATM Signaling
And Routing
ATM
Physical
Edge Device (a)
IP(a), ATM(a)
ATM Signaling
And Routing
ATM
ATM
Switch (c)
ATM(c)
ATM overlay model (Source: Cisco Systems)
OSPF,
NLSP, etc
ATM Signaling
And Routing
ATM
ATM
Switch (d)
ATM(d)
ATM Signaling
And Routing
ATM
Physical
Edge Device (b)
IP(b), ATM(b)
Internetworking
• Network layer
– Responsible for end-to-end connectivity
– Based on logical and topological addressing
– Addressing at Layer 3 is protocol dependent
• Routing
– Routing protocols: protocol specific or suitable for two or more
internetwork protocols (IP, IPX, …)
Application QoS and resource reservation
• Integrated Services Internet model
• Resource Reservation Protocol (RSVP)
• IP Precedence
The Integrated Services (IS) Internet model
• Real-Time (RT) QoS with control over end-to-end delay
• Support for multicasting
• Ability to assign percentages of a physical link bandwidth to different
traffic classes
Traffic Classes
Real-Time
Intolerant
Guaranteed
Tolerant
Controlled Load
Elastic
Interactive
Burst
Interactive
Bulk
Best Effort
Asynchronous
Bulk
The integrated Services (IS) Internet model
• Support of different QoS for real-time applications in addition to the
traditional Best Effort service
• Traffic control
– Packet scheduler – determines which packets will be transferred to the
physical medium and in what order
– Classifier – maps incoming packets into a given class for equal
treatment by the packet scheduler
– Admission control – determines whether or not a new flow may be
established without affecting existing sessions
– The ability to mark some packets as eligible for discard under network
congestion
Resource ReSerVation Protocol (RSVP)
• RSVP reservation
– Flow spec – application QoS requirements (bandwidth, delay)
– Filter spec – determines to witch packets the flow spec will apply
• The protocol operates in simplex mode
• RSVP-enabled network
– Sources characterize their flows by generating PATH (TSPEC)
messages, transmitted downstream (source to receivers) along the data
packet route as determined by the unicast or multicast routing protocol
– A receiver will generate an RESV (RSPEC) message, that is sent hopby-hop upstream to the unicast address of the previous RSVP hop
– The RESV message is used by the sender to set first hop traffic
parameters
IP precedence
• IP header contains IP precedence bits
• This field has existed since the first deployment of IP
• Represents different levels of priority for user traffic (IP Class of
Service)
– i.e. 2 - standard traffic, 4 - premium traffic
IP QoS to ATM QoS interworking
Layer 3
Peer 1
ATM
Layer 3
Peer 2
-Weighted Fair Queuing
(WFQ) on the packet side
of the ATM interface
One VC – maximum service guarantee
-One VP at maximum QoS
Layer 3
Peer 1
ATM
-Two VPs
-one for real-time traffic
- one for best-effort
One VC for each priority
Layer 3
Peer 2
IP QoS to ATM QoS interworking
Integrated service Guaranteed
specification
Controlled load
Best effort
End-to-End
behavior
Guaranteed
max delay
Best-effort on
unloaded net
Best-effort only
Intended
applications
Real-time
Sensitive to
congestion
Legacy
Control
mechanism
Leaky bucket, Leaky bucket
Reserved rate
and WFQ
ATM mapping
CBR, rt-VBR
None
nrt-VBR, ABR w/ UBR, ABR
MCR
Integrated Service Specifications
•Applicable to RSVP
•Applicable to IP Precedence by choosing proper values
Deploying QoS and the ATM Programming Interface
• API – map application QoS requirements to ATM Virtual Channel
Connection (VCC) parameters
Applications
Requests
Integrated
Services Internet
RSVP
PIM
ATM
Signaling
VC
Routing
Guarantees
Flow Specs Flow IDs
Packet
Scheduling
API
ATM
Traffic
Contract
VPI / VCI
Traffic
Mgmt
WinSock 2
• Independent of a specific Layer 3 protocol or any Layer 2 network
• Supports QoS, allows applications to utilize both pt-pt and pt-mpt
VCCs, and uses ATM addressing
• QoS support within WinSock 2 is based on the flow specification
– Source traffic description (token bucket size and token rate)
– Latency (time elapsed between a bit sent by the sender and arrival at
the destination), delay variation (the difference between the minimum
and maximum latency)
WinSock 2
PC Windows implementation
Application
WinSock 2
IPX
SNA NetBeui TCP/IP
Direct
Protocol
Access
Eth./TRN
LAN
Emul.
Clas.
IP
Direct
AAL
access
Signaling
ATM Adaptation Layer (AAL)
Ethernet &
Token Ring
Mac Driver
Eth. / TRN
Adapter
ATM Layer
Physical Layer
155 Mbps ATM | 25.6 ATM
MMF (fiber)
UTP
STP
Classical IP and ARP over ATM
• ATM network acting as a replacement for existing Layer 2
• ATM cloud treated as a single or as multiple Logical IP Subnets
(LISs)
• Classical IP and ARP over ATM (Classical model)
– Routers to connect members of different LISs
– Address resolution over the ATM network (ATMARP) and Inverse ATM
Address resolution protocol (InATMARP) – for only a single LIS
• Classical IP and ARP WAN over ATM
– ATM as a core network (subnetwork)
– Users, each on separate IP networks, will interconnect via the ATM
network
Multicasting and broadcasting
Differences between internetwork and ATM multicast
Internetwork multicasting
ATM multicasting
Bidirectional
Unidirectional
Multipoint-to-Multipoint
Point-to-Multipoint
Receiver-initiated join
Sender-initiated join (UNI 3.0/3.1)
Leaf-initiated join (UNI 4.0)
Connectionless
Connection oriented
Open-group (non-members may send) No multicast ‘group’ concept at
present
Multicasting and broadcasting
• Two new entities within an ATM network
– Multicast Address Resolution Server (MARS)
– Multicast Server (MCS) (optional) – used for mpt-mpt sessions
• Broadcasting
– A special case of multicasting
– A group containing all hosts
– The MARS maintain the registry of all group members, in this case all
end systems within the LIS
Optimizing routing
• Next Hop Resolution Protocol (NHRP)
– Extension of the Classical model and the Non-Broadcast Multi-Access
(NBMA) Address Resolution Protocol (NARP)
– An IP source station will use NHRP to determine the best IP and link
layer (NBMA) address to use to reach a destination station
ATMARP
Next Hop Servers
NS2
NS3
NS1
NS4
NH-Reply
NH-Requests
LIS1 LIS2
LIS3
LIS4
LIS1
LIS2
LIS3
Direct Connection
4 VCCs
Source (S)
Single VCC
Destination (D)
LIS4
IPv6
IP Version 6 Header:
Version
Flow Label
Payload Length Type
Hops
Source Address
(16 Octets)
Destination Address
(16 Octets)
Options
•Unicast, Multicast and Broadcast support
Requirements:
Real Time support
Hierarchical Addressing and
IPv4 Compatibility;
Autoconfiguration
Source Routing, Authentication,
Encryption, etc