Soft Examination Chapter 6 Network Interconnection and Internet

Networking and the Internet

1. Network interconnection equipment

The various networks that make up the Internet are called subnets, and the devices used to connect the subnets are called intermediate systems. Its main function is to coordinate the work of various networks, so that cross-network communication can be realized.

Network interconnection devices can be classified according to the protocol layers they work on:

• Repeater: works on the physical layer
• Bridges and switches: work at the data link layer
• Router: works at the network layer
• Gateway: A protocol that works above the network layer

1.1 Repeater

Connecting two network segments with a repeater can extend the transmission distance of the signal. The function of the repeater is to regenerate and send the received signal. Limit the use of a maximum of 4 repeaters in Ethernet. The repeater works on the physical layer and only plays the role of extending the transmission distance, which is transparent to high-level protocols. Repeaters can also connect networks of different transmission media together.

The working principle of a hub is the same as that of a repeater. Simply put, a hub is a multi-port repeater that broadcasts data received on one port to all other ports.

1.2 Bridge

The bridge analyzes the frame address field to decide whether to forward the received frame to another network segment. The bridge works on the MAC sublayer, as long as the protocols above the MAC sublayer of the two networks are the same, they can be interconnected through the bridge. For example, the interconnection between Ethernet over coaxial cable and Ethernet over twisted pair, or between Ethernet and token ring network can be realized.

1.3 Router

Routers pass packets between interconnected subnets based on network logical addresses. Routers are suitable for connecting complex large-scale networks. They work at the network layer, so they can be used to connect networks that implement different protocols in the lower three layers, and the protocol conversion is completed by routers.

1.4 Gateway

A gateway is the most complex network interconnection device, which is used to connect subnets that implement different protocols on the network layer to form a heterogeneous Internet. The gateway can convert mutually incompatible high-level protocols. In order to realize the communication between heterogeneous devices, the gateway needs to translate and transform different transport layer, session layer, presentation layer and application layer protocols. Due to the complexity of the work, the efficiency of using the gateway to the Internet is relatively low, and the transparency is not good, so it is often used for a dedicated connection for a special purpose.

2. WAN interconnection

WAN interconnection is generally realized by performing protocol conversion at the network layer, and the interconnection device used here is a router.

2.1 OSI network layer internal structure

OSI divides the network layer into three sublayers:

• Subnet access sublayer: corresponds to the actual network layer, which may or may not conform to the OSI network layer standard. If the subnet access sublayers of two actual networks are different, they cannot be simply interconnected

• Subnet related sublayer: enhance the service of the actual network, making it close to the network layer service of OSI

• Subnet-independent sublayer: Provides standard OSI network services.

2.2 Connection-Oriented Internetwork

The X.75 gateway implements the X.25 protocol to realize the interconnection of two subnets. An Internet virtual circuit is established between the two networks.

2.3 Connectionless Internetwork

The Internet Protocol (IP) is an Internet datagram protocol developed for ARPAnet, and later ISO developed the Connectionless Network Protocol (CLNP) based on it. The IP protocol is described here, and CLNP is similar to IP.

3. IP protocol

IP addresses are divided into 5 categories. Class A, B, and C addresses are used in point-to-point communication, and Class D addresses are multicast addresses. Class E addresses are reserved for research purposes, and future IPv6 addresses will be expanded on this basis.

IP datagram:

4.ICMP

ICMP and IP belong to the same network layer and are used to transmit messages about communication problems, such as datagrams that cannot reach the destination station, routers that do not have enough buffer space, or routers that provide shortest path information to the assistant. ICMP messages are encapsulated and transmitted in IP datagrams, so reliable delivery is not guaranteed.

5. TCP and UDP

There are two transmission protocols in the TCP/UDP protocol cluster, namely Transmission Control Protocol (TCP) and User Datagram Protocol (UDP). TCP is connection-oriented, while UDP is connectionless.

5.1 TCP protocol

Port numbers are used to identify TCP users:

application layer protocol	The port number
FTP	20 (data) + 21 (control)
SSH secure login, file transfer and port redirection	22
Telnet remote login protocol, insecure text transfer	23
SMTP Simple Mail Transfer Protocol	25
DNS	53
HTTP	80
POP3 Post Office Protocol Version 3	110
IMAP Internet Mail Access Protocol	143
HTTPS	443

Three handshakes and four waves:

5.2 TCP Congestion Control

Calculate the smoothed round-trip time estimate:
$$SRTT(K+1)=a\times SRTT(K)+(1-a)\times RTT(K+1)$$
Setting of retransmission timer:
$$RTO(K+1)=min(UPBOUND,max(LOWBOUND,b\times SRTT(K+1)))$$
slow start and congestion control:

For details on the TCP part, please refer to the top-down method

5.3 UDP protocol

The UDP protocol is divided into a header field and a data field. The header field only occupies 8 bytes, which are source port, destination port, length and checksum which occupy two bytes.

6. Gateway protocol

6.1 Autonomous System

An autonomous system is connected to the Internet by gateways of the same configuration, and the gateways within the autonomous system execute the Interior Gateway Protocol (IGP) to exchange routing information with each other. Exterior Gateway Protocol (EGP) exchanges routing information between different autonomous systems.

6.2 Exterior Gateway Protocol

The latest Exterior Gateway Protocol is called BGP, and now BGP4 has been widely used between networks of different ISPs, and has become the de facto Internet Exterior Routing Protocol standard. BGP4 is a dynamic route discovery protocol that supports CIDR. The main function of BGP is to control the routing strategy, such as whether to forward the passing packets. BGP is based on TCP connections.

BGP running within the same AS is called iBGP, and BGP running between different ASs is called eBGP.

Summary of BGP protocol

Four types of BGP messages:

message type	Functional description
open	establish neighbor relationship
update	Send new routing information
keep alive	Reply to open / Periodically confirm neighbor relationship
notification	Report detected errors

In BGP, the above four types of messages can be used to realize three functional processes:

• Establish a neighbor relationship: open–keep alive
• Neighbor reachability: periodically send keepalives to each other
• Network reachability: Each router maintains a database of all subnets it can reach. When the situation changes, the update message is used to transmit the latest information to other BGP routers in time. Information includes:
  • Network Layer Reachability Information (NLRI). Send a list of subnet addresses that the router can reach.
  • Autonomous systems traversed (AS_path).
  • Next Hop (Next-Hop)

6.3 Interior Gateway Protocol

Common interior routing protocols include Routing Information Protocol (RIP), Open Shortest Path First (OSPF), Intermediate System to Intermediate System Protocol (IS-IS), Interior Gateway Routing Protocol (IGRP), and Enhanced IGRP Protocol (EIGRP), etc. . The last two are Cisco's proprietary agreements.

RIP

Adopts Bellman-Ford's distance vector routing algorithm, which is suitable for small networks because it allows no more than 15 hops. RIP uses hop technology to measure routing costs.

RIPv2 packets are encapsulated in UDP datagrams and sent, occupying port number 520.

OSPF

Is a link-state protocol for exchanging routing information between routers within an autonomous interior. OSPF collects link information from each router, constructs a network topology diagram, and uses Dijkstra's shortest path first algorithm to calculate the best route to each target. OSPF uses a TCP connection to send packets, and each packet requires a response, making communication more reliable.

In order to meet the needs of large-scale network configuration, OSPF protocol introduces the concept of "hierarchical routing".

6.4 Core Gateway Protocol

Core Gateway Protocol (GGP): The core gateway in the backbone network is directly controlled by InterNOC, and GGP is more professional. When a core gateway joins the backbone network, the GGP protocol is used to broadcast routing information to neighboring machines, and each neighboring machine updates the routing table and further propagates new routing information.

7. Router technology

7.1 NAT technology

Mainly to solve the problem of IP address shortage, NAPT also performs port number conversion.

7.2 CIDR technology

Representation and division method of CIDR IP address

7.3 Layer 3 switching technology

Multi-protocol Label Switching (MPLS) developed by IETF integrates the link state of layer 2 into the protocol data unit of layer 3, thus simplifying and improving the switching process of layer 3 packet switching. When a packet enters the MPLS network, the Label Edge Router (LER) adds a label to it. This label not only contains the information in the routing table entry, but also refers to the source address field and transport layer port number in the IP header. and service quality etc. Once this classification is established, the packet is assigned to the corresponding label switching path (LSP), and the label switching router (LSR) will process the packet according to the label, and will not be forwarded through the third layer, thus speeding up the transmission speed of the network.

8. IP multicast technology

Some people describe the IP multicast model as: you inject packets at one end, and the network can just submit the packets to any desired receivers. The effectiveness of multicast technology lies in: when a multicast group is submitted to all multicast members, only the intermediate nodes related to the group can copy the group, and only one copy of the group is transmitted on the network link leading to each group member , so the use of multicast technology can improve the efficiency of network transmission and reduce the possibility of backbone network congestion. The prerequisite for implementing IP multicast is that the underlying network between the multicast source and group members must support multicast.

8.1 Internet Group Management Protocol

IGMP is a protocol that provides group management in the IPv4 environment. Multicast hosts and routers use IGMP to exchange multicast membership information to support hosts to join or leave a multicast group. In the IPv6 environment, the group management protocol has been merged into the ICMPv6 protocol, and a separate group management protocol is no longer needed. IGMP messages are encapsulated in IP datagrams for transmission.

Membership inquiry messages are sent by multicast routers and are divided into 3 subtypes:

• Generic queries: used by a router to know which groups are members of the network it is connected to
• Group-specific queries: Used by a router to know if a specific group has members on the network it is connected to
• Group and Source Specific Queries: Used by a router to know if a host it is connected to would like to join a particular group

In order to maintain a list of currently active multicast addresses, multicast routers periodically send IGMP general inquiry messages, encapsulated in IP datagrams with 224.0.0.1 (all hosts) as the destination address. A host that still wishes to maintain membership in one or more groups must read this datagram and reply with a report message for the groups it maintains membership in.

A multicast router does not need to know the address of each host in the multicast group. For a multicast group, it only needs to know that at least one multicast member is active. So members can wait for others to post (

When a host wants to leave a group, it sends a group leaving report to all routers. When a router receives such a report, it needs to determine whether there are other members of the group. At this time, the group and source-specific inquiry report can be used. arts.

8.2 Multicast Routing Protocols

Establishing a multicast tree is a key technology for realizing multicast transmission.

• Dense mode routing protocol: It is assumed that multicast members are densely distributed throughout the network, and the network has sufficient bandwidth, allowing periodic flooding to establish and maintain distribution trees.
• Sparse mode routing protocol: suitable for Internet networks with small bandwidth and sparse distribution of multicast members.

9. IP QoS technology

Best-Effort service model: best effort

Best-Effort is the simplest QoS service model. Users can send any number of packets at any time without notifying the network. When providing the Best-Effort service, the network sends packets as much as possible, but does not provide any guarantee for performance such as delay and packet loss rate. The Best-Effort service model is suitable for services that do not require high performance such as delay and packet loss rate. It is the default service model of the Internet and it is suitable for most network applications, such as FTP and E-Mail.

IntServ service model: reserved resources

The IntServ model means that users need to describe their traffic parameters to the network through signaling (Signaling) and apply for specific QoS services before sending packets. Based on traffic parameters, the network reserves resources to promise to satisfy the request. After receiving the confirmation information and confirming that the network has reserved resources for the application's message, the user starts to send the message. The packets sent by users should be controlled within the range described by traffic parameters. Network nodes need to maintain a state for each flow, and perform corresponding QoS actions based on this state to meet the commitment to users.

The IntServ model uses the RSVP (Resource Reservation Protocol) protocol as signaling to reserve resources such as bandwidth and priority on a known path network topology, and each network element along the path must Reserving desired resources, through the reservation of RSVP information, each network element can judge whether there are enough resources to use. Only when all network elements provide enough resources for RSVP, can a "path" be established.

DiffServ Service Model: Differentiated Services

The basic principle of the DiffServ model is to divide the traffic in the network into multiple classes, and each class enjoys different processing, especially when the network is congested, different classes will enjoy different levels of processing, thus obtaining different packet loss rates and delays and delay jitter. Services of the same type will be aggregated and sent uniformly in the network to ensure the same QoS indicators such as delay, jitter, and packet loss rate.

In the Diffserv model, the classification and aggregation of service flows are completed by border nodes at the edge of the network. Border nodes can flexibly classify packets according to various conditions (such as source and destination addresses of packets, priority in ToS field, protocol type, etc.), set different tag fields for different packets, and other Nodes only need to simply identify these tags in the message to perform resource allocation and flow control.

Compared with the Intserv model, the DiffServ model does not require signaling. In the DiffServ model, before the application program sends a message, it does not need to apply for resources to the network in advance, but informs the network node of its QoS requirements by setting the QoS parameter information of the message. The network does not need to maintain state for each flow, but provides differentiated services based on the QoS parameter information specified for each packet flow, that is, divides the service level of packets, performs flow control and forwarding differently, and provides end-to-end QoS guarantee. The DiffServ model fully considers the flexibility and scalability of the IP network itself, and converts the complex quality of service guarantee into a single-hop behavior through the information carried in the message itself, thereby greatly reducing the signaling work. mainstream service model.