CSIS 6251 CSIS 625 Week 10 Networking & Internetworking Copyright Dan Oelke For use by students of CSIS 625 for purposes of this class only.

CSIS 6252 Overview Networking and Internetworking –Moving to layer 3 –Routing types –Routing algorithms –Multicast Transport Layer –Up to layer 4

CSIS 6253 Vocabulary internet - a collection of networks connected by bridges or routers Internet - THE world wide collection of networks using TCP/IP as their network protocol that people use for communicating. End System (ES) - a node on one of the networks in an internet that supports end-user services (OSI model terminology) Intermediate System (IS) - a node that connects two networks to permit communications between end systems on the different networks. (OSI model terminology) IP - Internet Protocol - the network layer protocol used on the Internet.

CSIS 6254 Vocabulary Router - a device that uses the network layer information for forwarding packets from one network to another. Bridge - a device that uses data link layer information for forwarding packets from one network to another. Gateway is kind of like a router in that it connects multiple networks - but it does so at the application layer, instead of at the network layer. Repeater - a device that connects multiple network segments at the physical layer –Converts analog to digital, retimes & reshapes signal, convert back to physical layer transmission scheme Amplifier - A device that increases the amplitude of a analog signal fed into it.

CSIS 6255 Network Layer Provide a link between networks –These networks may be of different data link and/or physical layers Provide routing and delivery of data between nodes on different networks When there are many networks connected in multiple ways, the algorithms to determine how a packet gets from end to end get “interesting”

CSIS 6256 Connection oriented vs. connectionless Connection Oriented –Assumed that all networks provide a connection oriented form of service –May be virtual circuits on the networks –IS systems splice together connections between the networks. –This system is rarely used in real systems Connectionless –packet-switching instead of virtual circuits –At each router a forwarding decision is made independently for each packet

CSIS 6257 Routing Characteristics Routing algorithms and protocols are supposed to get packets from one node to another. How well they do this is judged on a number of factors –Correctness – packets get where they are supposed to –Simplicity – Able to implement this (simple makes equipment cheaper) –Robustness – Being able to deal with network problems –Stability – Making sure that things don’t change too rapidly causing more problems –Fairness – everyone gets the same amount (or what they paid for) –Optimality – bandwidth of links is well used. –Efficiency – make sure that processing is minimal so that delay is minimal

CSIS 6258 Least Cost Routing The goal of most routing protocols is to get information between two points in the “best” way. Best may be defined by a number of things –Number of network hops (easy to measure) –Amount of delay from various links Can be physical delay from transmission rate and distance Can add in congestion (queuing time) –Cost in $ to send packets over a given link –Usually is not physical distance

CSIS 6259 Adaptive vs. Non-adaptive routing Adaptive Routing –Each router as it processes each packet makes a decision about how to send the packet to its destination –This can change when the network changes Link failures Congestion Non-Adaptive routing –Once a pathway is established, all packets for a destination go along that one route.

CSIS Fragmentation and Reassembly Network layer (as well as other layers sometimes) provides segmentation and reassembly. Makes bigger packets of data into smaller ones that the underlying layer can handle. Each header has fields –Length –Offset value –“More” flag Example:

CSIS Packet Time to Live Packets may end up in a routing loop going around and around –May be just bounced between two nodes To keep packets from using network bandwidth forever, most network protocols have a packet lifetime specified by the originator. TTL – Time to live. - A number set by the packet originator and decremented by each hop along a path. –When this counter reaches 0, the packet is discarded –Used to limit the damage of routing loops.

CSIS Gateway A gateway is kind of like a router in that it connects multiple networks - but it does so at the application layer, instead of at the network layer. A special kind of application that transfers information from one application format to another.

CSIS Routing algorithms – in abstract There are many different ways to determine the best path for a packet to take through a network. Routing algorithms are the steps taken to find the best path Routing protocols are a description of how this routing information is discovered and disseminated in the network. Type of Routing Algorithms –Fixed Routing –Flooding –Random Routing –Distance Vector Routing –Link State Routing

CSIS Fixed Routing A simple method where a human goes to each router and programs it with tables that tell it where every packet goes. –May have a central network controller that disseminates the information Very simple and stable. Does not react well to network congestion or link failures. –May have alternate paths for each destination to accommodate link failures.

CSIS Flooding Routing Very simple method where every node sends packets to every other node it is connected to. Must have a mechanism to kill off packets –Nodes could remember every packet it has sent –A TTL counter in the packet can be implemented Multiple copies will be received by the recipient so packet must have unique tag that allows duplicates to be discarded.

CSIS Flooding Advantages: –Requires no central authority. –All links are tried – packets will get through if there is any way possible. Very robust – good for emergency messages in a military network. –At least one packet will have used minimum hop count May be used to find path for virtual circuit –All nodes receive the packet May be used to disseminate important information (like route updates) Disadvantage –Very high network load for the traffic given Actually used in some routing protocols and in the peer-to- peer application Gnutella.

CSIS Random Routing Send a packet to one random outgoing path for retransmission. Same idea as flooding, but with less traffic load on the network. Advantages –No central authority –Relatively robust –Less traffic than flooding Disadvantages –Still a heavy traffic load –Most packets do not use the least hop path.

CSIS Distance Vector Routing A type of adaptive routing Each router periodically shares its knowledge about the entire network. –This is sent only to the router’s direct neighbors. –This information is shared at a regular basis When a router receives information from its neighbor, it updates its routing table. –The routing table has Network ID, cost, next hop. –When a lower cost path is found, the old route is discarded and the new route added.

CSIS Link State Routing A type of adaptive routing Each router shares its knowledge about it’s neighbors (not the entire routing table) Information about it’s neighbors is sent to all routers –Uses a flooding technique Information is sent out when there is a change (not periodically) When a router receives information, it uses it to update its routing table –The routing table has Network ID, cost, next hop.

CSIS Dijkstra Algorithm To calculate the the lowest cost path between two nodes, the routers use Dijkstra’s algorithm. The algorithm builds a tree structure of the network using itself as the root. –All nodes that can be reached from the root are attached (all neighbors) – temporarily. –The node are sorted by order of cost to reach them. –Starting with lowest cost temporarily attached node, make it permanent part of the tree. –Consider all nodes attached from the chosen node and add them temporarily. –Repeat last two steps until all nodes are attached permanently Example:

CSIS Routing algorithms – in practice Vocabulary –Autonomous system A group of networks and routers where all the routers exchange information using a common routing protocol. All of these routers are managed by a single organization Except where there is a failure, all routers are “connected” –IRP – Interior Router Protocol – A routing protocol used within a single autonomous system. Also known as Intra-Domain Routing protocol –ERP – Exterior router protocol – A routing protocol used for exchanging routing information outside of an autonomous system. Also known as Inter-Domain Routing Protocol Typically an ERP is simpler than an IRP –Exchanges only summary information of reachability –IRP has more detailed information on least-cost path to reach any given node.

CSIS BGP – Border Gateway Protocol An Exterior router protocol. Designed to allow routers (“gateways” in the standard) of different autonomous systems to exchange information. BGP-4 Defined in RFC major functions/procedures –Neighbor acquisition –Neighbor reachability –Network reachability Distributes information for each neighbor –List of routers needed to get to the neighbor –IP address of the router that is the next hop –List of networks served by this router

CSIS RIP – Routing Information Protocol An early TCP/IP routing protocol. Now used as an Interior routing protocol only Depreciated in use for the most part –Since it was one of the first, it still shows up in a lot of systems. Each router broadcasts it’s entire route table. A Distance Vector routing protocol Has problem’s scaling as the number of routers and links grows very large

CSIS OSPF – Open Shortest Path First An upgrade from RIP An Interior routing protocol Based on Link State Routing Cost to traverse a link may be set to anything that the network administrator desires. –May be delay, data rate, $, etc. –Some “costs” may be artificially inflated, or reduced to help steer traffic down a certain path.

CSIS IS-IS Routing An OSI stack based system – now adopted for TCP/IP networks

CSIS Multicast Traffic Multicast - to send the same data to multiple destinations, but not send multiple copies and not broadcast it to everyone. Useful for: –Radio/TV broadcasts where users “tune in” –Teleconferencing – IETF meetings are often sent this way –Distributed updates of information (software updates, database updates, etc) Uses special set of network (and sometimes data link layer) addresses. On a single broadcast LAN, often sent as a broadcast to a special address that allows network interfaces to listen (or ignore) as they choose On some systems that are not multicast aware, it can be sent multiple times – also called multiple unicast.

CSIS Special requirements of multicast Routers must be multicast aware. Router will possibly forward a packet out multiple ports rather than just one. Each multicast aware router must keep track of networks or interfaces that have are “joined” to a particular multicast session. Routers must handle nodes, or networks, leaving and joining a multicast session. Routing protocol and algorithms needed for routers to determine the shortest path to all group members.

CSIS Transport Layer May be connection oriented (TCP) or connectionless (UDP) Connection Oriented Transport protocol –Provides establishment, maintenance, and termination of a logical connection End to end delivery of messages (not just packets) –Provides segmentation and reassembly of messages into packets Addressing - addition of port number Flow Control Ordered Delivery Reliable Delivery Duplicate Detection

CSIS Transport Layer Transport layer provides for reliable delivery –At least the TCP part of TCP/IP does –There is UDP/IP which is not reliable Error control and flow control typically done using a sliding window mechanism. –Sequence numbers with ACKs and NAKs Transport provides for connection establishment and termination –A 3 way handshake is typical for connection establishment. –Obviously, not needed for connectionless protocols

CSIS Transport Layer - Port numbers Transport Layer adds to network address the SAP – Service Access Point –In TCP/IP and many protocols this is called the port number –Provides an additional level of addressing beyond the host. Allows for an additional level of multiplexing –Typically identifies the service – HTTP server SMTP server POP3 server Telnet server etc

CSIS Port numbers How does a user application know what port number to use? –User “just knows” the number - it is a configuration option –Well known port numbers are used /etc/services on many systems This is commonly used for servers –A name server is used –Another application on a well-known port spawns a child application on some other port (remote job management)

CSIS Connection Establishment Typically a three-way handshake Initiator sends a SYN (Synchronize sequence number) packet Receiver sends back a SYN packet that also acknowledges the initiators initial sequence number Initiator sends an ACK packet to acknowledge the receiver’s initial sequence number Now either side may start sending data If the SYN packets overlap - no problem both just send ACKs.

CSIS Connection Establishment Security concerns The initial sequence number must be random to prevent session hijacking. –If not, a malicious sender can create packets that look like they come from a trusted source and inject any data that they choose. A malicious initiator can send a lot of initial SYN packets, but never finish the 3-way transaction –This can cause resources on the receiver to be tied up until the three-way handshake times out.

CSIS Connection Termination One side decides it is done and sends a FIN (Finish) packet to the other. The other side responds with a FIN packet. After receiving the corresponding FIN packet back the session is considered closed. If you receive a FIN packet, it is considered closed after sending a FIN packet back.

CSIS Sequence numbers Some systems use a sequence number per packet. Some systems use an implicit sequence number for each byte. –This means that sequence numbers can increase a bunch for every packet of data. –TCP uses this scheme By ensuring sequence numbers occur in order we get: –Ordered delivery –Error control for lost or damaged packets –Flow Control –Duplicate detection

CSIS Retransmission strategy A positive acknowledgement of each received segment is required If an acknowledgement is not received after some time period, a retransmission of the segment occurs –May be lost data segment -or- lost ACK Timeout for retransmission –May be a fixed value - but it is difficult (impossible?) to get a good value for all situations Too long means sluggish response to lost packets Too short means many retransmissions for packets that were delayed (not lost) Ideal timer is just a little longer than round-trip time –May be adaptive Difficult because transmission and processing delays can change widely and rapidly.

CSIS Duplicate detection & Out of Order Data Management A receiver doesn’t know if a duplicate is the first copy or second –The first copy may have been delayed causing the second copy to arrive before the first. The receiver acknowledges the first copy received The sequence number window must be large enough so that a packet will die before sequence numbers wrap around If data is received out of order –Receiver may discard segment –Receiver may hold segment and wait for missing segment