Outline

network layer
virtual circuit and datagram networks
datagram networks
forwarding design
IP
IP fragmentation

TCP Summary

TCP connections: state at the endpoints (real connections involve the network as well)
TCP state machine: closed, to established, to closed
three-way handshake
closing connections, 2x half-closing
sequence numbers, acks, timers for reliable transmission
careful timer design to retransmit as soon as possible, but no sooner
multiple (3) duplicate acks used as a NAK: fast retransmission, also fast recovery for congestion control
congestion "window" maintained by sender, AIMD for discovery of available throughput
flow control window sent (in acks) by receiver

Network Layer

transferring a packet from an incoming link to an outgoing link of a router is forwarding
getting a packet from source to destination, perhaps over many routers, is routing
when a packet reaches a router, its destination address (or other value in the header) is used as an index in a table, the routing table or forwarding table
each row of the table has (at least) a destination address and an output link
a routing protocol is used to build and maintain the routing table
the routing protocol implements a routing algorithm
a router is an example of a packet switch
a router is a packet switch that forwards based on the network-layer header
another example is an ethernet switch, which forwards based on the link-layer header

Network-Layer Connections

the Internet is connectionless
but other network-layer technologies (ATM, Frame Relay) are connection-oriented
in a connection-oriented network, the network layer must also provide connection management

Network Service Models

network-wide: guaranteed delivery, possibly with bounded delay
per-flow:
- in-order delivery
- guaranteed minimum bandwidth
- guaranteed maximum jitter
- security: encryption, authentication
Internet: best-effort (none of the above)
ATM Available Bit Rate (ABR): in-order delivery, minimum bandwidth guarantee, congestion notification
ATM Constant Bit Rate (CBR): in-order delivery, minimum bandwidth guarantee, maximum jitter guarantee, maximum packet loss

Virtual Circuit Network

the Internet Protocol is connectionless: data is forwarded based only on its destination address
other technologies, e.g. ATM, are connection-oriented: data is forwarded based on a connection identifier carried in each cell (packet)
a path through a network is a virtual circuit if, at each router/switch, it is identified by a Virtual Connection Identifier (VCI)
the forwarding table is indexed by interface and incoming VCI
an entry in the forwarding table must have outgoing interface and outgoing VCI
the VCI changes every time a cell is forwarded
the entries in the forwarding table must be managed by an appropriate signaling protocol which adds or removes entries in the forwarding table of each router/switch
the signaling protocol requires that routers communicate with each other to set up the virtual circuit
the granularity of Virtual Circuits can be set up so that each gets special treatment, e.g. higher priority, lower delay, lower chance of packet loss

Datagram Network

in a datagram network, forwarding for each packet is based on its destination address
since a destination address must be unique in the entire network, the routing table could grow very large

Datagram Network Routing Tables

to reduce the size of a routing table, it is good to summarize ranges of addresses and have the same route to all addresses in the range
this can be done if the initial part of an address, the prefix is assigned to one network, and routes direct packets to networks
the last router in a route, which is on the destination network, must know how to deliver the packets to the final destination
different prefixes may have different lengths
longer prefixes may match all the bits of another route that has a shorter prefix: the longer prefix represents a more specific route
so longer prefixes are preferred
in general, we need to record the prefix, the number of valid bits in the prefix, and the route, i.e. the outgoing interface
the prefix with zero bits is the least preferred, and holds the default route, used when no other route matches

Comparisons of Virtual Circuits to Datagrams

Virtual Circuits make it easier to provide specific services
Virtual Circuits are based on the true circuits of wired telephones
with Virtual Circuits, the complexity and cost are in the network
with datagrams, the complexity and cost are in the end-systems
in today's world, that means that datagram networks can be more flexible and provide new services faster
the network part of a datagram network is typically cheaper than the Virtual Circuit equivalent
the end-system part of a datagram network may have to be more complex than the equivalent in a Virtual Circuit network
but there are economies of scale to building such end-systems, so it need not be expensive

Fundamental Switch Design

see textbook figure 4.6 on p. 325
each input port needs an interface circuit (interface card) to translate bits-on-the-wire to packets
the interface card typically takes care of layers 1 (physical) and 2 (data link)
packets are handed to a lookup/forwarding unit that determines which output port(s) should get the packet
the switch control processor (routing processor) may also receive (or send) packets, e.g. to implement the routing protocol
there is a network inside the switch, which physically delivers the packets from the input port circuitry to the output port circuitry: the switching fabric
every output port needs a queue, and data link and physical layer circuitry
input and output ports are paired if links are bidirectional, but logically (and in the switching fabric) they are independent

A simple router

a regular PC with multiple interfaces can serve as a router:
the input and output port circuits are paired as a network interface card
each incoming packet is delivered to the main processor
forwarding/routing software, usually in the operating system (OS), determines what to do with the packet
outgoing packets are sent on the output port
if queueing is needed, it is often done in software, using a queue allocated in the main memory of the CPU by the OS

Switch Input Port Functions

in a hardware switch, no software is needed to forward packets
this gives higher performance compared with software routing
either the input port circuitry is responsible for selecting the output port, or this function may be built into the switching fabric
the routing processor must configure (program) each input port circuit (or the switching fabric) with the forwarding/routing table
many optimizations are used to achieve line speed processing of incoming packets, including:
- preprocessing the forwarding table for faster lookups
- caching recently addressed destinations
- content-addressable memories
the goal of the forwarding table lookup is to determine what output port (or ports, for multicast) the packet needs to go to

Switching Fabric Design

ideally, the switching fabric would be fast enough to carry all possible incoming packets
this is done sometimes, but is not always practical or cost effective
packets may be stored in memory along the way from input to output ports, e.g. in a general purpose computer used as a router
if the switching fabric is a bus, it interconnects all input to all output ports:
- signals on the bus must identify the output port(s), which will listen to this incoming packet
- the bus can only be used by one packet at a time, so
- either the bus is very fast, or packets may have to block temporarily
alternatives usually combine a number of buses with some way of transferring packets from one bus to another
the ultimate of these is a crossbar, where there is an input bus for every input port, capable of transferring packets to any of the output busses connected to each output port

Switch Output Port Functions

the output port manages the output queue, and the MAC and physical layer processing
the output queue must be designed to discard data in case of overflow

Queueing and Scheduling

queue sizing is still an area of research
suggestions vary: given N connections with round-trip time RTT, and link capacity C, a buffer size B = RTT * C is likely sufficient, but B = RTT * C / sqrt(N) might also be sufficient
often packets are forwarded in First-Come-First-Serve order
sometimes packets are forwarded according to priority, or to provide a certain Quality of Service (QoS) or in an attempt to maintain fairness
packets can also be dropped in order of arrival (drop tail, or drop the most recent arrival), or according to priority or QoS
managing the queue based on the packets it contains is called Active Queue Management, or AQM
Random Early Discard (RED) also Random Early Detection, is an AQM which discards TCP packets before the queue is full to ask the sending TCPs to slow down

Input Queues

a queue at an input port may be needed if the switching fabric is not fast enough to handle all possible incoming packets
in this case, depending on queue design, a packet for one output port may block another destined for an output port that could be reached
this is know as Head of Line (HOL) blocking
making the switching fabric fast enough avoids such problems and simplifies the design of the input ports

Internet Network Layer

the forwarding protocol in the network layer of the Internet is IP, the Internet Protocol
today, the commonly used version of IP is IPv4 (Internet Protocol version 4)
IPv6 is also widely available, but not (yet?) widely used
routing protocols are responsible for computing the forwarding tables (routing tables)
there are many routing protocols, mostly deployed in a hierarchical fashion
the network layer of the Internet also includes a protocol for error reporting and information reporting, the Internet Control Message Protocol or ICMP

Internet Protocol

IPv4 is defined by RFC 791, with a major update in RFC 1122, with a major

each IP packet is sent with an IP header (from RFC 791):

    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 
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |Version|  IHL  |Type of Service|          Total Length         |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |         Identification        |Flags|      Fragment Offset    |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  Time to Live |    Protocol   |         Header Checksum       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                       Source Address                          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                    Destination Address                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                    Options                    |    Padding    |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

the basic IP header is 20 bytes, and always a multiple of 4 bytes

IP header fields

the version number is 4

the Internet Header Length (IHL) is the length of the header in 4-byte, and has a minimum value of 5

the type of service can be used to provide differentiated service to different packets

the last two bits of the type of service field are used to mark packets for Explicit Congestion Notification

the length includes header and data

identification, flags, and fragment offset are used to fragment packets where needed, then reassemble them at the destination

the time to live (TTL) field is decremented each time the packet is forwarded (or stored for at least a second), and the packet is dropped if the TTL reaches zero

the protocol field identifies the higher-layer protocol:

1 means this is an ICMP packet
6 means this is a TCP packet
17 (0x11) means this is a UDP packet

the header checksum is the Internet Checksum computed over all the bytes of the header

the source and destination address identify the sender and the intended receiver of the message

these fields may be followed by options, then usually by data

Internet Header Sizes

a TCP ack packet has a TCP header and an IP header: if there are no options, a TCP ack is 40 bytes
all other TCP packets are more than 40 bytes
a UDP packet has a 20-byte IP header, an 8-byte UDP header, and usually at least 1 byte of data, for a total of 29 bytes

IP fragmentation

each link-layer protocol has a maximum size packet it will carry: the maximum transmission unit or MTU
a router, or even an IP host sending a UDP packet, may need to send IP packets larger than the MTU
since the Internet is designed to accomodate many different technologies, the only hard limit on packet size is given by the Total Length field, which is 16 bits
all technologies that can support IP are required to have an MTU of no less than 576 bytes
an IP router receiving a packet too large for the MTU can:
- drop the packet, if the DF (Don't Fragment) flag is set, and send an ICMP error message to the original sender, or
- split the packet into smaller fragments, each of which is an IP packet, and send the fragments to the destination
a destination receiving an IP fragment must:
- recognize that this is a fragment
- put the fragment's data in the proper place in the reconstructed packet
- discard the packet if the fragment is missing for long enough
routers are not supposed to reassemble fragments

Fragmentation Details

a sender uses a different Identification value for each packet sent to each receiver, so the receiver can distinguish fragments belonging to different IP packets
each fragment specifies the position of this fragment's data in the reassembled packet, the Fragment Offset
the 13-bit Fragment Offset field actually stores 1/8 of the actual fragment offset
the flags are DF (Don't Fragment, discussed earlier) and MF, More Fragments
MF is set in every fragment, except the last fragment
a receiver getting a packet:
- with Fragment Offset = 0 and MF = 0 knows this packet is not fragmented
- with Fragment Offset > 0 and MF = 0 knows the length of the final reassembled packet
- with Fragment Offset = p and MF = 1 knows this is a fragment whose data belongs at offset 8p within the final reassembled packet
fragments can be fragmented again (by a subsequent router) if necessary

example IP headers and fragmentation

the header of a 1500-byte UDP packet:
```
45 00 05 DC
49 41 00 00
15 11 12 34
80 AB 03 0D
80 AB 01 01
```
(the checksum is likely not correct)
this packet has ID 0x4941, or 18753
this packet has 1480 bytes of IP payload
to forward this packet on a network with MTU 576, we split it into two payloads of size 552 (0x228) and one payload of size 376 (0x178)
note payloads for all but the last fragment should have sizes that are multiples of 8
as always, the TTL is decremented in the outgoing headers

the three headers are:

45 00 02 40
49 41 20 00
14 11 12 34
80 AB 03 0D
80 AB 01 01

45 00 02 40
49 41 20 45
14 11 12 34
80 AB 03 0D
80 AB 01 01

45 00 01 8C
49 41 00 8A
14 11 12 34
80 AB 03 0D
80 AB 01 01