Professional Documents
Culture Documents
Shivkumar Kalyanaraman
Overview
Switched Packet-Data Services Integrated Services Vision and Concept Ingredients History: X.25, ISDN, Frame Relay ATM Networks: foundation for B-ISDN ATM Key Concepts ATM Signaling and PNNI Routing ATM Traffic Management IP over ATM: setting the stage for MPLS
Shivkumar Kalyanaraman
Separate Voice network (PSTN) and Data Networks (Frame Relay, SMDS, etc.) PSTN sometimes used as a data network backbone, but PSTN is circuit switched (voice-optimized) and PSTNbased WAN not efficient Delay sensitive traffic such as voice not possible on data networks since no guarantee of QoS Initial attempts to converge data and voice network not too successful, i.e. ISDN B-ISDN and ATM networks viewed as the convergence end-point leading world-wide domination of telephony driven standards
Shivkumar Kalyanaraman
After the success of T1, the telephone carriers saw the growth in packet switched networks Evolved their own flavors of packet switching, notably X.25, ISDN, SMDS, Frame Relay, ATM etc Key concept: Switched services Switched services: (aka dial-up service) Digital communications that is active only when the customer initiates a connection. Subsumes both circuit switched and packet switched. Customer to be billed only when the line is active. Led to activity-based or average-load-based pricing models that did not necessarily have a distance-based component Vs peak-rate and distance-sensitive T-carrier pricing
4
Shivkumar Kalyanaraman
Ingredients
Signaling and setup of a virtual circuit (I.e. nailing down a switched path) is a common feature Signaling was heavyweight, and was coupled to heavyweight QoS routing Contrast this to connectionless, best-effort Internet Long 20-byte global addresses used only in signaling Short 4-byte local labels (aka DLCI etc) used in packets (cells): label-switching Large address space, low per-packet overhead
ISDN/B-ISDN vision of an end-to-end integrated digital network: Rich QoS capabilities developed: support for voice, data, Institute video traffic Shivkumar Kalyanaraman Rensselaer Polytechnic
Ingredients (contd)
X.25 -> Frame relay/ATM: reduction of hop-by-hop processing complexities Led to the development of high-speed switches and networks A serious attempt to inter-network with a variety of data-networking protocols (IP, Ethernet etc) Integration (coupling) of too many features led to slow rollout, enormous overall complexity Failure to attain the end-to-end market vision Current trend is to de-couple building blocks of the architecture within the context of IP/MPLS, sacrificing strict performance guarantees.
Shivkumar Kalyanaraman
X.25
Shivkumar Kalyanaraman
X.25
First packet switching interface in the telephony world Issued in 1976 and revised in 1980, 1984, 1988, and 1992. Data Terminal Equipment (DTE) to Data Communication Equipment (DCE) interface User to network interface (UNI) Slow speeds, used in point-of-sale apps (eg: credit-card validation) and several apps abroad
Shivkumar Kalyanaraman
Circuit: Pin a path, reserve resources, use TDM based transmission Virtual Circuit = Virtual Call: pin a path, optionally reserve resources Connection-oriented: Setup an end-to-end association (datastructure); path not pinned Connectionless: stateless. No path, no end-to-end association Two Types of Virtual Circuits: Switched virtual circuit (SVC): Similar to phone call Permanent virtual circuit (PVC): Similar to leased lines Up to 4095 VCs on one X.25 interface
Shivkumar Kalyanaraman
Note: the three modular layers were co-specified by the same standards body Layers: X.21 replaced by EIA-232 (RS-232C) LAP-B = Link access procedure - Balanced Packet layer = Connection-oriented transport over virtual circuits
Shivkumar Kalyanaraman
10
Electrical and mechanical specifications of the interface X.21 = 15-pin digital recommendation X.21bis = X.21 twice = X.21 second Interim analog specification to allow existing equipment to be upgraded. Now more common than X.21 => X.21 Rev 2
RS-232-C developed by Electronics Industries Association of America (EIA) is most common Uses 25-pin connector. Commonly used in PCs.
Shivkumar Kalyanaraman
11
Original: Synchronous Data Link Control (SDLC): IBM Derivatives: High-Level Data Link Control (HDLC): ISO Link Access Procedure-Balanced (LAPB): X.25 Link Access Procedure for the D channel (LAPD): ISDN Link Access Procedure for modems (LAPM): V.42 Point-to-Point Protocol (PPP): Internet Logical Link Control (LLC): IEEE
Link Access Procedure for half-duplex links (LAPX): Teletex Advanced Data Communications Control Procedures (ADCCP): ANSI V.120 and Frame relay also use HDLC
Shivkumar Kalyanaraman
12
HDLC (contd)
Primary station: Issue commands (master) Secondary Station:Issue responses (slave) Hybrids: Combined Station: Both primary and secondary: a.k.a Asynchronous Balanced Mode (ABM) Balanced Configuration: Two combined stations Unbalanced Configuration: One or more secondary Normal Response Mode (NRM): Response from secondary Asynchronous Response Mode (ARM): Secondary may respond before command
Shivkumar Kalyanaraman
13
LAPB
Uses balanced mode subset of HDLC between DTE and DCE Uses 01111110 as frame delimiter Uses bit stuffing to avoid delimiters inside the frames Uses HDLC frame format Point-to-point: Only two stations - DTE (A), DCE (B) Addresses: A=00000011, B=00000001 Address = Destination Addresses in Commands
Shivkumar Kalyanaraman
14
HDLC frames
Information Frames: User data Piggybacked Acks: Next frame expected Poll/Final = Command/Response Supervisory Frames: Flow and error control Go back N and Selective Reject Final No more data to send Unnumbered Frames: Control Mode setting commands and responses Information transfer commands and responses Recovery commands and responses Miscellaneous commands and responses
Shivkumar Kalyanaraman
15
HDLC Operation
SABM: Set Asynchronous Balanced Mode UA: Unnumbered ACK DISC: disconnect RR: Receiver Ready RNR: Receiver Not Ready I: information frame Heavyweight Link-Setup and Per-Packet Acking !!
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
16
Shivkumar Kalyanaraman
17
Packet Level = End-to-end for X.25 networks But really Layer 3 (network layer) Packet level procedures: Establishment and clearing of virtual calls Management of PVCs Flow Control Recovery from error conditions
Shivkumar Kalyanaraman
18
Shivkumar Kalyanaraman
19
GFI = Packet formatting information PTI = 20 possible packet types (for de-multiplexing)
Logical Channel Group and Channel Numbers: Virtual circuit identifier
Shivkumar Kalyanaraman
20
Fragmentation/Reassembly support: M = More segments Layer 3 reliability: P(R) and P(S) refer to packet sequence # Different from N(R) and N(S) - frame sequence #
Shivkumar Kalyanaraman
21
3-bit and 7-bit sequence number options possible Again, note: these are layer 3 sequence numbers
Shivkumar Kalyanaraman
22
Shivkumar Kalyanaraman
23
Shivkumar Kalyanaraman
24
ISDN Configurations
Shivkumar Kalyanaraman
25
* Basic Rate ISDN and Primary Rate ISDN. * BRI can transmit data up to 128 kbps. * PRI (transmitted over a T1 line) can transmit data up to 1.536 Mbps.
An LDN (Local Directory Number): customer's 7-digit ISDN phone number. A SPID (Service Profile Identifier): unique ID of an ISDN line or service provider (10+ digits long and includes the LDN).
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
26
Basic Rate ISDN service divides a standard telephone line into three digital channels capable of simultaneous voice and data transmission. The three channels are comprised of two Bearer (B) channels at 64 kpbs each and a data (D) channel at 16 kbps, also known as 2B+D. The B channels are used to carry voice, video, and data to the customer's site (hence the term integrated services). The D channel is used to carry signaling and supplementary services. Multiple B channels can be used at the same time. The D channel can also be used to carry packetized data.
Shivkumar Kalyanaraman
27
Shivkumar Kalyanaraman
28
U-interface: U-interface is a 2-wire digital telephone line that runs from the telephone company's central office to an NT1 device.
NT1 (Network Termination Type 1): NT1 is a Basic Rate ISDN-only device that converts a service provider's U-interface to a customer's S/T-interface. Stand-alone or integrated into a terminal adapter.
S/T-interface: S/T-interface is a common way of referring to either an S- or T-interface. This can be used to connect directly to an ISDN 2B+D NT1 or an NT2 device with a terminal adapter. This type of interface is often found on Terminal Equipment Type 1. TE1: TE1 (Terminal Equipment Type 1) is ISDN-ready equipment that can directly connect to the ISDN line (often using an S/ Tinterface). Eg: ISDN phones, ISDN routers, ISDN computers, etc.
Shivkumar Kalyanaraman
29
TA (terminal adapter): TA is a device that allows nonISDN-ready equipment to connect to an ISDN line. This device can have an integrated NT1. R-interface: R-interface is a non-ISDN interface such as an EIA-232 or a V.35 interface. This type of interface is often found on TE2. TE2 (Terminal Equipment Type 2): TE2 is equipment that cannot directly connect to an ISDN line. A common example of this device is a PC, or a non-ISDN-ready router. A TA must be used to connect to the ISDN line.
Shivkumar Kalyanaraman
30
Primary Rate Interface (PRI) ISDN is a user-to-network interface (UNI) consisting of: Twenty-three 64 kbps bearer (B) channels, and One 64 kbps signaling (D) channel (aka 23B+D) Cumulatively carried over a 1.544 Mbps DS-1 circuit. The B channels carry data, voice or video traffic. The D channel is used to set up calls on the B channels.
Shivkumar Kalyanaraman
31
Shivkumar Kalyanaraman
32
Shivkumar Kalyanaraman
33
Shivkumar Kalyanaraman
34
Frame Relay
Shivkumar Kalyanaraman
35
Multiple logical links => Multiple connections Four nodes => 12 ports (full mesh!!) 12 local exchange carrier (LEC) access lines, 6 inter-exchange carrier (IXC) connections One more node => 8 more ports, 8 more LEC lines, 4 more IXC circuits (same issues as full mesh in LANs) Charged both by bandwidth and by the mile!
Shivkumar Kalyanaraman
36
6 IXC circuits (star vs full mesh: FR network is like a hub or switch in a star-topology) One more node: 1 more port, 1 more access line, 4 more IXC circuits Share local leased lines to LECs (aka Virtual Private Networks (VPNs) or closed-user groups (CUGs)) Tradeoffs: Packetized L2 (FR) or L3 (X.25) service instead of digital L1 service (T-carrier) Service guarantees weaker (delay, jitter, loss; PIR/CIR vs peak rate)
Shivkumar Kalyanaraman
37
FR obviously more efficient from a protocol standpoint than X.25, in addition to the compelling economics vs leased lines
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
38
X.25: interface between host and packet-switching network 3 layers: phy, link, packet Heavyweight: error control at every link as well as layer 3: twelve messages for one packet transfer!! X.25 offers no QoS capability
Frame relay breaks up link-layer into two parts: LAPF-core and LAPF-control Network nodes only implement LAPF-core Frame Switching is a service that implements both
Frame relay uses a separate VC for control channel in vs in-band control approach used in X.25
Shivkumar Kalyanaraman
39
Frame Relay: digital packet network providing benefits dedicated T-1 link, but without the expense of multiple dedicated circuits. Frame Relay leverages the underlying telephone network Frame Relay distance-insensitive and average-rate pricing is an ideal, cost-effective solution for networks with bursty traffic
Especially those that require connections to multiple locations and where a certain degree of delay is acceptable. FR also allows a voice circuit to share the same virtual connection as a data circuit, again, saving money.
Switches do not perform hop-by-hop error correction (other than discarding corrupted frames) or flow control (other than setting FECN/BECN bits) Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
40
X.25 simplified No flow and error control Out-of-band signaling Two layers Protocol multiplexing in the second layer Congestion control added Higher speed possible. X.25 suitable to 200 kbps vs Frame relay suitable to 2.048 Mbps. Frame Relay = Unreliable multiplexing service X.25 Switching = Relaying + Ack + Flow control + Error recovery +loss recovery
Shivkumar Kalyanaraman
41
PVC: Permanent Virtual Circuit DLCI: Data Link Connection Identifier CIR: Committed Information Rate CSU: Channel Service Unit UNI: User-to-Network Interface NNI: Network-to-Network Interface DTE: Data Terminal Equipment DE: Discard Eligible FRAD: Frame Relay Access Device DSU: Data Service Unit Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
42
Frame Relay Access Device FRAD: generic name for a device that multiplexes/formats traffic for entering a Frame Relay network. Access Line: A communications line interconnecting a Frame Relay-compatible device to a Frame Relay switch. Bursty/burstiness: Sporadic use of bandwidth that does not use the total bandwidth of a circuit 100% of the time. CIR (Committed Information Rate): The committed rate (usually < the access/peak rate) which the carrier guarantees to be available DE (Discard Eligibility): A user-set bit: frame may be discarded DLCI (Data Link Connection Identifier): A unique number IDing a particular PVC endpoint: has local significance only to that channel. BECN (Backward Explicit Congestion Notification): A bit set by a FR network to notify an interface device (DTE) that congestion avoidance procedures should be initiated by the sending device. FECN (Forward Explicit Congestion Notification): A bit set by a FR network to notify an interface device (DTE) that congestion avoidance procedures should be initiated by the receiving device.
Shivkumar Kalyanaraman
43
DTE (Data Terminal Equipment): User terminal equipment which creates information for transmission; for example, a user's PC or a router. CSU/DSU: A customer owned, physical layer device that connects DTE (eg: router) to an access line (eg: T1), from the network service provider. Traditionally, DSUs were network-owned equipment used in conjunction with customer-owned CSUs to terminate access lines. Because of regulatory changes, there is no need for physical separation of CSU and DSU any longer => combination CSU/DSUs.
Shivkumar Kalyanaraman
44
* Similar to X.25 DLCI: Only local significance * Multiple logical connections over one physical circuit * Some ranges pre-assigned Eg: DLCI = 0 is used for signaling
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
45
UNI = User-network Interface LAPF = Link Access Protocol - Frame Mode Services LAPD = Link Access Protocol - D Channel Control Plane: Signaling over D channel (D = Delta = Signaling) Data transfer over B, D, or H (B = Bearer) LAPD used for reliable signaling ISDN Signaling Q.933 + Q.931 re-used for signaling messages Service Access Point Identifier (SAPI) in LAPD = 0 => Q.933 + Q.931 Frame relay message Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
46
Functions:
Frame delimiting, alignment, and flag transparency Virtual circuit multiplexing and de-multiplexing Octet alignment => Integer number of octets before zero-bit insertion Checking min and max frame sizes Error detection, Sequence and non-duplication Congestion control
47
LAPF is similar to LAPD: Flag, bit stuffing, FCS No control frames in LAPF-Core => No control field No in-band signaling unlike X.25 No flow control, no error control, no sequence numbers Logical Link Control (LLC) may be used on the top of LAPF core
Shivkumar Kalyanaraman
48
Shivkumar Kalyanaraman
49
Minimum rate guarantee: Committed Information Rate (CIR) Maximum burst rate: Peak Information Rate (PIR) TM enforcement model: Discard Control (DE Bit) set on all packets when CIR < user rate < PIR Network usually over-provisioned for CIR, but underprovisioned for PIR Can drop packets with DE set during congestion (I.e. when absolutely necessary) Congestion control hooks:
Backward Explicit Congestion Notification (BECN) Forward Explicit Congestion Notification (FECN) Very nice ideas later proposed as ECN in TCP/IP But generally ignored in practice by CPE equipment
50
Shivkumar Kalyanaraman
Shivkumar Kalyanaraman
51
Shivkumar Kalyanaraman
52
Policing actions: Between Bc and Bc + Be => Mark DE bit Over Be => Discard
Shivkumar Kalyanaraman
53
FECN
Forward Explicit Congestion Notification (FECN) Source sets FECN = 0 Networks set FECN if avg Q >1 Dest tells source to inc/dec the rate (or window) Start with R = CIR (or W=1) If more than 50% bits set => decrease to 0.875 R (or 0.875W) If less than 50% bits set => increase to 1.0625 R (or min{W+1, Wmax}) If idle for a long time, reset R = CIR (or W=1)
Shivkumar Kalyanaraman
54
BECN
Backward Explicit Congestion Notification (BECN) Set BECN bit in reverse traffic or send Consolidated LinkLayer Management (CLLM) message to source On first BECN bit: Set R = CIR On further "S" BECNs: R=0.675 CIR, 0.5 CIR, 0.25 CIR On S/2 BECNs clear: Slowly increase R = 1.125 R If idle for long, R = CIR
Shivkumar Kalyanaraman
55
BECN (Contd)
For window based control: S = One frame interval Start with W=1 First BECN W = max(0.625W,1) Next S BECNs W = max(0.625W,1) S/2 clear BECNs => W = max(W+1, Wmax) CLLM contains a list of congested DLCIs
Shivkumar Kalyanaraman
56
Shivkumar Kalyanaraman
57
Driven by the integration of services and performance requirements of both telephony and data networking broadband integrated service vision (B-ISDN) Telephone networks support a single quality of service and is expensive to boot Internet supports no quality of service but is flexible and cheap ATM networks are meant to support a range of service qualities at a reasonable cost Intended to subsume both the telephone network and the Internet
Shivkumar Kalyanaraman
58
ATM Concepts
1. Virtual circuits 2. Fixed-size packets (cells): allowed fast h/w switching 3. Small packet size 4. Statistical multiplexing 5. Integrated services 6. Good management and traffic engineering features 7. Scalability in speed and network size Together can carry multiple types of traffic with end-to-end quality of service
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
59
ATM Applications
ATM Deployments: Frame Relay backbones Internet backbones Aggregating Residential broadband networks (Cable, DSL, ISDN) Carrier infrastructures for the telephone and privateline networks Failed market tests of ATM: ATM workgroup and campus networks ATM enterprise network consolidation End-to-end ATM
Shivkumar Kalyanaraman
60
Phone networks are synchronous (periodic). ATM = Asynchronous Transfer Mode Phone networks use circuit-switching. ATM networks use Packet or cell Switching In phone networks, all rates are multiple of 64 kbps. With ATM service, you can get any rate, and you can vary your rate with time. With current phone networks, all high speed circuits are manually setup. ATM allows dialing any speed & rapid provisioning
61
Shivkumar Kalyanaraman
ATM is virtual circuit based: the path (and optionally resources on the path) is reserved before transmission
Internet Protocol (IP) is connectionless, and end-to-end resource reservations not possible RSVP is a new signaling protocol in the Internet IP packets: variable size Internet provides best-effort routing (combination of RIP/OSPF/IS-IS/BGP-4), aiming only for connectivity ATM uses 20-byte global NSAP addresses for signaling and 32bit locally-assigned labels in cells IP uses 32-bit global addresses in all packets TCP/IP: congestion control is packet-loss-based
62
Addressing:
Shivkumar Kalyanaraman
Shivkumar Kalyanaraman
63
ATM Interfaces
UNI = User-Network Interface (Private & Public) NNI = Network Node Interface (Private and Public) B-ICI = Broadband Inter-Carrier Interface DXI = Data Exchange Interface Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
64
Shivkumar Kalyanaraman
65
Shivkumar Kalyanaraman
66
ATM Layers
Adaptation: mapping apps (eg: voice, data) to ATM cells Physical layer: SONET etc ATM Layer: Transmission/Switching/Reception, Congestion Control, Cell header processing, Sequential delivery etc
Shivkumar Kalyanaraman
67
AAL Sublayers Convergence Sublayer (CS) Determines Class of Service (CoS) for incoming traffic Provides a specific AAL service at an AAL network service access point (NSAP) Segmentation and Reassembly Sublayer (SAR) Segments higher-level user data into 48-byte cells at the sending node and reassembles cells at receiving node Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
68
AAL Lingo.
Shivkumar Kalyanaraman
69
AAL Types
Shivkumar Kalyanaraman
70
Transports ATM cells on a communications channel and defines mechanical specs (connectors, etc.) 2 Sub-layers Transmission Convergence Sub-layer Maps cells into the physical layer frame format (e.g. DS1, STS3) on transmit and delineates ATM cells in the received bit stream Generates HEC on transmit Generates idle cells for cell rate decoupling, or speed matching Physical Medium Sub-layer Medium dependent functions like bit transfer, bit alignment, OEO
Shivkumar Kalyanaraman
71
Physical Layers
Multimode Fiber: 100 Mbps using 4b/5b, 155 Mbps SONET STS-3c, 155 Mbps 8b/10b Single-mode Fiber: 155 Mbps STS-3c, 622 Mbps Plastic Optical Fiber: 155 Mbps Shielded Twisted Pair (STP): 155 Mbps 8b/10b Coax: 45 Mbps, DS3, 155 Mbps Unshielded Twisted Pair (UTP) UTP-3 (phone wire) at 25.6, 51.84, 155 Mbps UTP-5 (Data grade UTP) at 155 Mbps DS1, DS3, STS-3c, STM-1, E1, E3, J2, n T1
Take-home message: Serious attempt to inter-operate with several L1, L2 and L3 technologies Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
72
ATM-SONET Mapping
Cells are mapped row-wise into the frame Cells could contain data or be empty
Rensselaer Polytechnic Institute
Shivkumar Kalyanaraman
73
74
Two ways to use packets carry entire destination address in header carry only an identifier, a.k.a label Labels have local significance, addresses have global significance Signaling protocol: fundamentally maps global addresses or paths (sequence of addresses) to local labels Data VCI Addr. Data Data
75
Sample
ATM cell
Datagram
Shivkumar Kalyanaraman
All packets must follow the same path (why?) Switches store per-VCI state: eg: QoS info Signaling => separation of data and control Small Ids can be looked up (exact match) quickly in hardware harder to do this with IP addresses (longest-prefix match) Setup must precede data transfer delays short messages Switched vs. Permanent virtual circuits
Shivkumar Kalyanaraman
76
ATM Switches
Shivkumar Kalyanaraman
77
Shivkumar Kalyanaraman
78
Shivkumar Kalyanaraman
79
Pros Simpler buffer hardware packet arrival and departure requires us to manage fixed buffer sizes Simpler line scheduling each cell takes a constant chunk of bandwidth to transmit Easier to build large parallel packet switches Cons overhead for sending small amounts of data segmentation and reassembly cost last unfilled cell after segmentation wastes bandwidth
Shivkumar Kalyanaraman
80
Shivkumar Kalyanaraman
81
Shivkumar Kalyanaraman
82
Trade off worst-case delay against speed of output trunk Whenever long term average rate differs from peak, we can trade off service rate for delay
Build scheduling, buffer management, policing entities to manage the zero-sum games of delay and bandwidth
Shivkumar Kalyanaraman
Internetwork or WAN
Data Plane: Traffic conditioning (shaping, policing, marking etc) + Traffic Classification + Scheduling, Buffer management
Shivkumar Kalyanaraman
84
ABR (Available bit rate): Source follows network feedback. Max throughput with minimum loss. UBR (Unspecified bit rate): User sends whenever it wants. No feedback. No guarantee. Cells may be dropped during congestion. CBR (Constant bit rate): User declares required rate. Throughput, delay and delay variation guaranteed. VBR (Variable bit rate): Declare avg and max rate. rt-VBR (Real-time): Conferencing. Max delay guaranteed. nrt-VBR (non-real time): Stored video.
Shivkumar Kalyanaraman
85
Shivkumar Kalyanaraman
86
Classes of Service
The Convergence Sublayer (CS) interprets the type and format of incoming information based on 1 of 4 classes of service assigned by the application
Class A: Constant bit rate (CBR), Connection oriented, strict timing relationship between source and destination, i.e voice Class B: Variable bit rate (VBR), Connection oriented, strict timing, e.g. packet-mode video for video conferencing Class C: Connection oriented VBR, not strict timing, e.g. LAN data transfer applications such as Frame Relay Class D: Connectionless VBR, not strict timing, e.g. LAN data transfer applications such as IP
Shivkumar Kalyanaraman
87
ABR vs UBR
ABR Queue in the source Pushes congestion to edges Good if end-to-end ATM Fair Good for the provider UBR Queue in the network No backpressure Same end-to-end or backbone Generally unfair Simple for user
Shivkumar Kalyanaraman
88
Shivkumar Kalyanaraman
89
Shivkumar Kalyanaraman
90
Shivkumar Kalyanaraman
91
Shivkumar Kalyanaraman
92
Shivkumar Kalyanaraman
93
Shivkumar Kalyanaraman
94
Shivkumar Kalyanaraman
95
Shivkumar Kalyanaraman
96
Shivkumar Kalyanaraman
97
Shivkumar Kalyanaraman
98
Shivkumar Kalyanaraman
99
Three NSAP-like (Network Service Access Point) address formats: DCC ATM Format, ICD ATM Format, Polytechnic E.164 ATM Format Shivkumar Kalyanaraman Rensselaer Institute
100
Shivkumar Kalyanaraman
101
Flat addresses: no structure in them to facilitate scalable routing Eg: IEEE 802 LAN addresses Hierarchical addresses: Network part (prefix) and host part Helps identify direct or indirectly connected nodes
Shivkumar Kalyanaraman
102
Authority and Format Identifier (AFI) & IDI: 39 = ISO DCC, 47 = British Stds Institute ICD, 45 = ITU ISDN ISDN uses E.164 numbers (up to 15 BCD digits) ATM forum extended E.164 addresses to NSAP format. E.164 number is filled with leading zeros to make 15 digits. A F16 is padded to make 8 bytes. End System Identifier (ESI): 48-bit IEEE MAC address. Selector is for use inside the host and is not used for routing. All ATM addresses are 20 bytes long.
103
Shivkumar Kalyanaraman
NSAP = Network Service Access Point. Identifies network layer service entry SNPA = Sub-network point of attachment. Identifies the interface to sub-network SNPA address (or part of it) is used to carry the packet across the network. CLNP uses NSAP to deliver the packet to the right entity in the host. ATM uses NSAP-like encoding but ATM addresses identify SNPA and not NSAP.
Shivkumar Kalyanaraman
104
Permanent and Switched Point to point Symmetric or asymmetric bandwidth (Uni- or bidirectional) Point-to-multipoint: Data flow in one direction only. Data replicated by network. Leaf Initiated Join (LIJ) or non-LIJ
Shivkumar Kalyanaraman
105
Shivkumar Kalyanaraman
106
Shivkumar Kalyanaraman
107
Shivkumar Kalyanaraman
108
Shivkumar Kalyanaraman
109
Shivkumar Kalyanaraman
A hierarchy mechanism ensures that this protocol scales well for large world-wide ATM networks. A key feature of the PNNI hierarchy mechanism is its ability to automatically configure itself in networks in which the address structure reflects the topology
Shivkumar Kalyanaraman
111
PNNI Features
Scales to very large networks. Supports hierarchical routing. Supports QoS. Supports multiple routing metrics and attributes. Uses source routed connection setup. Operates in the presence of partitioned areas. Provides dynamic routing, responsive to changes in resource availability. Separates the routing protocol used within a peer group from that used among peer groups. Interoperates with external routing domains, not necessarily using PNNI. Supports both physical links and tunneling over VPCs.
Shivkumar Kalyanaraman
112
Peer group: A group of nodes at the same hierarchy Border node: one link crosses the boundary Logical group node: Representation of a group as a single point Child node: Any node at the next lower hierarchy level Parent node: LGN at the next higher hierarchy level Logical links: links between logical nodes Peer group leader (PGL): Represents a group at the next higher level.
Node with the highest "leadership priority" and highest ATM address is elected as a leader. PGL acts as a logical group node. Uses same ATM address with a different selector value.
Shivkumar Kalyanaraman
PNNI Terminology
Shivkumar Kalyanaraman
114
Shivkumar Kalyanaraman
115
Shivkumar Kalyanaraman
116
Shivkumar Kalyanaraman
117
Shivkumar Kalyanaraman
118
Source Routing
Source specifies route as a list of all intermediate systems in the route (original idea in token ring) Designated Transit List (DTL): (next slide) Source route across each level of hierarchy Entry switch of each peer group specifies complete route through that group Set of DTLs and manipulations implemented as a stack
Shivkumar Kalyanaraman
119
DTL Example
Shivkumar Kalyanaraman
120
If a call fails along a particular route: It is cranked back to the originator of the top DTL The originator finds another route or Cranks back to the generator of the higher level source route
Shivkumar Kalyanaraman
121
Shivkumar Kalyanaraman
122
Connection Admission Control (CAC): Can requested bandwidth and quality of service be supported? Traffic Shaping: Limit burst length. Space-out cells. Usage Parameter Control (UPC): Monitor and control traffic at the network entrance. Network Resource Management: Scheduling, Queueing, virtual path resource reservation Selective cell discard: Cell Loss Priority (CLP) = 1 cells may be dropped Cells of non-compliant connections may be dropped Frame Discarding Feedback Control: ABR schemes
Shivkumar Kalyanaraman
123
Shivkumar Kalyanaraman
124
Peak Cell Rate (PCR): 1/T Sustained Cell Rate (SCR): Average over a long period Burst Tolerance (BT) ts : GCRA limit parameter wrt SCR GCRA(1/Ts, ts) Maximum Burst Size: MBS = 1+BT/(1/SCR-1/PCR) BT [(MBS-1)(1/SCR-1/PCR), MBS(1/SCR- 1/PCR)] Cell Transfer Delay (CTD): First bit in to last bit out Cell Delay Variation (CDV): ~ Max CTD - Min CTD Peak-to-peak CDV Cell Delay Variation Tolerance (CDVT) t = GCRA limit parameter wrt PCR GCRA(T, t)
Cell Loss Ratio (CLR): Cells lost /Totals cells sent Minimum cell rate (MCR)
125
Shivkumar Kalyanaraman
Peak-to-Peak CDV
Shivkumar Kalyanaraman
126
Service Categories
Shivkumar Kalyanaraman
127
Provides traffic shaping: I.e. smooth bursty arrivals Provides traffic policing: Ensure that users are sending traffic within specified limits Excess traffic discarded or admitted with CLP = 1 GCRA in ATM requires increment (inter-cell arrival time) and limit (on earliness) Two implementations: Virtual scheduling and leaky bucket
128
Shivkumar Kalyanaraman
Shivkumar Kalyanaraman
129
Shivkumar Kalyanaraman
130
Shivkumar Kalyanaraman
131
GCRA: Examples
Shivkumar Kalyanaraman
132
Shivkumar Kalyanaraman
133
DECbit scheme in many standards since 1986. Forward explicit congestion notification (FECN) in Frame relay Explicit forward congestion indicator (EFCI) set to 0 at source. Congested switches set EFCI to 1 Every nth cell, destination sends an resource management (RM) cell to the source Shivkumar Kalyanaraman Rensselaer Polytechnic Institute
134
Shivkumar Kalyanaraman
135
Shivkumar Kalyanaraman
136
Shivkumar Kalyanaraman
137
IP OVER ATM
Shivkumar Kalyanaraman
138
Shivkumar Kalyanaraman
139
One ATM LAN can be n virtual LANs Logical subnets interconnected via routers Need drivers in hosts to support each LAN Only IEEE 802.3 and IEEE 802.5 frame formats supported. (FDDI can be easily done.) Doesn't allow passive monitoring No token management (SMT), collisions, beacon frames. Allows larger frames.
Shivkumar Kalyanaraman
140
LAN Emulation driver replaces Ethernet driver and passes the networking layer packets to ATM driver. Each ATM host is assigned an Ethernet address. LAN Emulation Server translates Ethernet addresses to ATM addresses Hosts set up a VC and exchange packets All software that runs of Ethernet can run on LANE
Shivkumar Kalyanaraman
141
Shivkumar Kalyanaraman
142
Shivkumar Kalyanaraman
143
Terminology
NDIS = Network Driver Interface Specification ODI = Open Datalink Interface IPX = NetWare Internetworking Protocol LAN Emulation Software: LAN Emulation Clients in each host LAN Emulation Servers LAN Emulation Configuration server (LECS) LAN Emulation Server (LES) Broadcast and unknown server (BUS)
Shivkumar Kalyanaraman
144
Shivkumar Kalyanaraman
145
LANE Process
Address Resolution: Client sends ARP request to Server. Unresolved requests sent to clients, bridges. Server, Clients, Bridges answer ARP Client setups a direct connection Broadcast/Unknown Server (BUS): Forwards multicast traffic to all members Clients can also send unicast frames for unknown addresses
Shivkumar Kalyanaraman
146
Shivkumar Kalyanaraman
147
IP over ATM
How many VCs do we need for n protocols? Packet encapsulation [RFC1483] How to find ATM addresses from IP addresses Address resolution [RFC1577] How to handle multicast? [MARS, RFC 2022] How do we go through n subnets on a large ATM network? [NHRP]
Shivkumar Kalyanaraman
148
Shivkumar Kalyanaraman
149
Question: Given an ATM link between two routers,how many VCs should we setup? Answer 1: One VC per Layer 3 protocol. Null Encapsulation: No sharing. VC based multiplexing.
Shivkumar Kalyanaraman
150
Answer 2: Share a VC using Logical Link Control (LLC) Subnetwork Access Protocol (SNAP). LLC Encapsulation Protocol Types: 0x0800 = IP, 0x0806 = ARP, 0x809B = AppleTalk, 0x8137 = IPX
Shivkumar Kalyanaraman
151
IP address: 123.145.134.65 ATM address: 47.0000 1 614 999 2345.00.00.AA.... Issue: IP Address ATM Address translation Address Resolution Protocol (ARP) Inverse ATM ARP: VC IP Address Solution: ATMARP servers
Shivkumar Kalyanaraman
152
ATM stations are divided in to Logical IP Subnets (LIS) ATMARP server translates IP addresses to ATM addresses. Each LIS has an ATMARP server for resolution IP stations set up a direct VC with the destination or the router and exchange packets.
Shivkumar Kalyanaraman
153
Shivkumar Kalyanaraman
154
Routers assemble packets Slow NHRP servers can provide ATM address for the edge device to any IP host Can avoid routers if both source and destination are on the same ATM network.
Shivkumar Kalyanaraman
155
MPOA= LANE + NHRP+ Extension of LANE Uses NHRP to find the shortcut to the next hop No routing (reassembly) in the ATM network
Shivkumar Kalyanaraman
156
MPOA (contd)
LANE operates at layer 2 RFC 1577 operates at layer 3 MPOA operates at both layer 2 and layer 3 MPOA can handle non-routable as well as routable protocols Layer 3 protocol runs directly over ATM Can use ATM QoS MPOA uses LANE for its layer 2 forwarding
Shivkumar Kalyanaraman
157
Shivkumar Kalyanaraman
158