It has been too long since I last studied, so let me review. These are notes taken with reference to the study video

Huawei Data Communication

Network Models

OSI Model

1. Application Layer (Layer 7)

   • Provides interfaces and services to applications

2. Presentation Layer

   • Performs data format conversion, encryption and decryption, and image/text encoding

3. Session Layer

   • Establishes, maintains, and terminates communication sessions

4. Transport Layer

   • End-to-end transmission; responsible for ports and reliable/unreliable transmission

5. Network Layer

   • Routing and addressing; determines the data path

6. Data Link Layer

   • Transmission between adjacent devices, error checking, physical addressing

7. Physical Layer (Layer 1)

   • Physical hardware, signal transmission; only transmits the bit stream of 0/1

TCP/IP Model

1. Application Layer (Layer 4)

   • Combines the OSI application, presentation, and session layers. Faces users and programs

   • Protocols: HTTP (80), HTTPS (443), FTP (21), SSH (22), Telnet (23), SMTP (25), POM3 (110), IMAP (143), DNS (53), apps and browsers, etc.

2. Transport Layer

   • End-to-end transmission; distinguishes programs by port

   • Protocols: TCP (065535) reliable, UDP (065535) efficient

   • 0~1023 are well-known fixed ports, 1024+ are dynamic custom ports

3. Internet Layer (Internetwork Layer)

   • Routing and addressing; finds the destination host by IP address

   • Protocols: IP, ICMP

4. Network Interface Layer (Network Layer)

   • Combines the OSI data link layer and physical layer; responsible for hardware transmission, MAC addressing, and cable/signal forwarding

   • Devices: network interface cards, switches, cables

TCP/UDP Packet Format

Ethernet Frame Format

ARP Address Resolution Protocol

ARP Principle

1. HOST1 ARP cache

1. ARP request

1. HOST2 ARP cache — upon receiving, sends a unicast packet and saves it to the cache table

1. ARP response — upon receiving, sends a unicast packet and saves it to the cache table

1. HOST1 ARP response

IP Protocol

IPv4 Packet Format

Packet Fragmentation

1. The process of dividing a packet into multiple fragments

2. IP packets forwarded over the network can vary in length, but when a packet's length exceeds the maximum length of the data link layer, it must be divided into several smaller fragments for transmission

   • The data link layer has an MTU (Maximum Transmission Unit); Ethernet defaults to MTU=1500 bytes

Flag Bits

1. Identification

   • All fragments split from the same original packet have identical identification values; the receiver uses it to determine which fragments belong to the same large packet

2. Flags (3 bits, only the last 2 bits are used)

3. MF bit (More Fragment)

   • 1 = there are more fragments after this one

   • 0 = this is the last fragment

   • The other bit: the Don't Fragment bit. When set to 1, fragmentation is not allowed; an oversized packet is dropped directly and an ICMP error message is returned

4. Fragment Offset

   • 8 bytes

   • Indicates: how many 8-byte units the data portion of the current fragment is offset from the start of the original packet

   • Used for ordering and reassembly

   • Except for the last fragment, all other fragment lengths must be an integer multiple of 8 bytes

   • The fixed 20 bytes of the IP header (with no options field) must be excluded when calculating, i.e., 1480 bytes

5. Each fragment has its own IP header (the header is unchanged, only the data is split)

6. During routing and forwarding, fragments may travel over different paths and arrive out of order; the receiver relies on identification + fragment offset to sort and reassemble them

$\text{Fragment Offset} = \frac{\text{Total data bytes of all fragments before the current fragment}}{8}$

Time To Live (TTL)

1. An 8-bit field in the IPv4 header. Values: 0~255

2. Limits the maximum number of routing hops a packet can traverse; used to prevent loops

IP Classes

Class	Leading Bits	First Octet Decimal Range	Default Subnet Mask	Network Bits	Host Bits	Primary Use
A	0	1 ~ 126	255.0.0.0	8	24	Large networks
B	10	128 ~ 191	255.255.0.0	16	16	Medium networks
C	110	192 ~ 223	255.255.255.0	24	8	Small LANs
D	1110	224 ~ 239	—	—	—	Multicast
E	1111	240 ~ 255	—	—	—	Research / Reserved
Private IP addresses that can be freely used on the intranet, not consuming public addresses:

• Class A private range: 10.0.0.0 ~ 10.255.255.255

• Class B private range: 172.16.0.0 ~ 172.31.255.255

• Class C private range: 192.168.0.0 ~ 192.168.255.255

Routing

Route Acquisition Methods

1. Directly connected routes

   1. Automatically generated when devices are connected; priority 0, the highest

2. Static routes

   1. Manually configured; default priority 60

3. Dynamic routes

   1. Devices automatically exchange route information with each other

   2. Self-learning, self-updating

   3. OSPF routes

      1. Internal route priority 10

      2. External route priority 150

Static route configuration rule:

ip route-static {next-hop IP | outgoing interface | next-hop + outgoing interface}

Equal-Cost Routes

Definition: For the same destination network, multiple routes have exactly equal cost metrics, and the router forwards traffic across multiple links simultaneously
When the following three conditions are met, the routing table generates multiple route entries at the same time, enabling equal-cost load sharing. Prerequisites:

1. The destination IP network is the same

2. The cost calculated by the routing protocol is the same

3. The next hops and outgoing interfaces differ (two or more different physical paths exist)

Floating Static Routes

Definition: For the same destination network, two static routes are configured; the primary route has a higher priority, while the backup route with a lower priority does not enter the routing table. When the primary link fails, the backup automatically takes over
Key points:

1. Huawei / H3C: the smaller the priority value, the more preferred; Cisco is also the smaller the more preferred

2. Default static priority is 60

3. Floating backup route setting: 61~255 (the larger, the lower the priority)

OSPF

Definition

Definition: Open Shortest Path First, a classless inter-domain routing protocol, belonging to the IGP (Interior Gateway Protocol) and link-state routing protocols. Based on the SPF algorithm

1. Protocol number: 89

2. Multicast addresses: 224.0.0.5 (all OSPF routers), 224.0.0.6 (DR/BDR)

3. Features: classless, supports VLSM, supports route summarization, fast convergence, loop prevention; suitable for medium-to-large campus / carrier intranets

4. LSDB (Link-State DataBase): the "whole-network topology map" shared by all routers within the same OSPF area, storing all LSA entries of the area.

OSPF Principles

1. A link-state IGP protocol with IP protocol number 89; routers do not exchange routes directly with each other, but send each other LSA (link-state advertisement) messages describing links, costs, and neighbors

2. Routers within the same area fully synchronize an identical LSDB (link-state database)

3. Neighbors discover and maintain each other through Hello packets; after negotiation succeeds, LSAs are exchanged, and the LSDB reaches a Full adjacency

4. Each router runs the SPF algorithm rooted at itself, independently computing the shortest paths based on the LSDB to generate its routing table

5. Areas are divided into the backbone Area 0 and non-backbone areas; the purpose is to reduce the scope of LSA flooding; ABRs are responsible for inter-area route propagation

6. Broadcast networks elect DR/BDR to reduce the number of adjacencies

7. Cost is used as the metric, defaulting to 10^8/interface bandwidth

8. Different types of LSA handle inter-area, intra-area, and external routes; LSAs are flooded within their specified scope

The Functions and Establishment Process of OSPF's 8 States

State order: Down → Attempt → Init → 2‑Way → ExStart → Exchange → Loading → Full

Among them, Attempt appears only in NBMA networks (Frame Relay)

The complete interaction of the 8 states is as follows:

1. Down

   • If no Hello packet has been sent yet, it remains in the Down state

2. Attempt

   • Neighbors are manually configured; Hello packets are sent by unicast, waiting for a response from the peer

3. Init

   • After receiving a Hello packet, OSFP starts running and enters the Init state

   • Mainly to discover the peer and learn its link information, cost, neighbor information, etc.

   • For example: R1 sends a Hello packet to R2; R2 receives the Hello packet, sees the peer's Router-ID in it, and after learning the peer's relevant information, R2 replies with a Hello packet, writing R1's RID and related information into its neighbor list; R1 receives the packet, sees its own RID in the neighbor list, and enters 2-Way; at this point R1 also needs to send a Hello packet carrying R2's RID; R2 receives the packet, sees its own RID, and R2 enters 2-Way

   • This process requires both sides to receive each other's Hello packets; a bidirectional neighbor relationship is established

   • Router‑ID is just an identity; the neighbor list field is the credential for bidirectional confirmation.

4. 2-Way

   • Received a Hello packet, and the neighbor list in the Hello packet contains itself: bidirectional connectivity is established, and the neighbor relationship is formed.

   • After 2-Way, broadcast networks perform DR/BDR election; p2p networks skip it

   • Routers that are neither DR (Designated Router) nor BDR (Backup Designated Router) are called DRother

   • DRothers stay only at 2‑Way and do not proceed further

   • Devices that need to establish adjacency enter ExStart, e.g., DRother ↔ DR; DRother ↔ BDR

5. ExStart

   • After entering ExStart, LSDB synchronization begins; DBD packets are exchanged to elect the master/slave, typically the device with the larger Router‑ID becomes the master; it controls the DBD packet sequence numbers. Only master/slave is negotiated; no LSA summaries are exchanged.

   • After the master/slave is confirmed, it enters ExChange

6. ExChange

   • After entering ExChange, DBD packets carrying LSA summaries are exchanged. Each side tells the other which LSAs are in its LSDB; only LSA summaries are passed, not complete LSAs

7. Loading

   • Comparing DBD summaries, when a router finds it lacks certain LSAs, it starts sending LSR packets to request the complete LSAs

   • The peer replies to the LSR request with LSU packets (carrying the complete LSAs); after receiving them, it also replies with LSAck acknowledgments

   • Repeatedly request, update, and acknowledge

8. Full

   • When all LSAs are synchronized and the LSDB databases are consistent, the SPF algorithm starts running to compute routes, and the adjacency completes, entering the Full state

   • A complete OSPF routing table is generated. Note: each router has its own global routing table; neighbors do not pass route entries to each other, only LSA topology, and each computes its own routes. Among them, the LSA aging time is one hour

   • When there is no failure, Hello (10s) packets maintain the relationship; a Hello is sent every 10 seconds to ensure the neighbor is still alive

   • When four consecutive Hello packets go unanswered, there is a Dead Time of 40 seconds timer, i.e., the maximum allowed timeout for the neighbor: four Hello packet intervals. When the timer expires, the neighbor is judged to have failed; the neighbor and all LSAs learned from it are deleted, SPF computation restarts, and the state is reset to Down

   • When the network changes, an LSA with an updated version number is generated and flooded across the entire network using LSU packets; however, different LSA types have restricted propagation scopes, flooding only within the corresponding scope each LSA can reach — see the LSA classification for details

     • Routers receiving the LSU compare it against their LSDB, update their local LSDB, and reply with LSAck acknowledgment

     • Trigger SPF recalculation (not a full recomputation) to refresh the OSPF routing table

     • The neighbor state remains Full and does not fall back to ExStart; it only regresses when the neighbor fails and disconnects

OSPF Areas

The concept of area division: if the entire network is not divided and only Area 0 exists, all devices would share a single LSDB, making the network too large; it would also produce a huge number of LSAs, causing repeated full-network SPF runs and frequent network flapping

LSA Classification

Number	Common Name	LSA Name	Generator	Flooding Scope	Core Function
Type‑1	Type 1 LSA	Router LSA	Every router in the area	Local area	Describes all directly connected links and costs of the router; source of intra-area routes
Type‑2	Type 2 LSA	Network LSA	DR of the broadcast segment	Local area	Describes broadcast / NBMA multi-access segments, listing all routers on the segment
Type‑3	Type 3 LSA	Summary‑Network LSA	ABR	Other areas	Propagates routes between different areas, inter-area routes O IA; supports route summarization
Type‑4	Type 4 LSA	Summary‑ASBR LSA	ABR	Other areas	Advertises the Router‑ID location of the ASBR router, serving Type 5 external routes
Type‑5	Type 5 LSA	AS‑External LSA	ASBR	Entire OSPF AS	Introduces external routes (static, other protocols); E1/E2 two metrics
Type‑6	Type 6 LSA	Group‑Membership LSA	MOSPF router	Local area	Used by MOSPF multicast OSPF; not used by ordinary unicast OSPF
Type‑7	Type 7 LSA	NSSA‑External LSA	ASBR within the NSSA area	NSSA area only	NSSA area introduces external routes; NSSA‑ABR converts Type 7 to Type 5 LSAs; N1/N2
Type‑8	Type 8 LSA	External‑Attribute LSA	ASBR	Entire AS	Carries external route attributes such as BGP; barely used in live networks
Type‑9	Opaque LSA #9	Opaque‑LSA (link-local)	Router	Local link only	Extension features; cannot flood across routers
Type‑10	Opaque LSA #10	Opaque‑LSA (area scope)	Router	Local area	MPLS‑TE, Segment Routing; floods within the local area
Type‑11	Opaque LSA #11	Opaque‑LSA (entire AS)	Router	Entire AS domain	Extensions such as SR; floods across the whole OSPF domain

Backbone Area

Area 0 definition:

1. Exists and is unique; the backbone area

2. All normal areas must be directly connected to Area 0; areas cannot be daisy-chained onto other areas (Area 1 connecting to Area 2 without connecting to Area 0 is an incorrect topology)

3. Non-backbone areas that are isolated and cannot directly connect to Area 0 must configure a virtual link Virtual-link to reach the backbone

Area 0 function: after each area's routes are summarized, they are forwarded through the backbone, acting as the high-speed backbone of the entire OSPF network

Non-Backbone Areas

Areas 1 through N are non-backbone areas
Features: when the topology changes in this area, LSAs do not leave the area, and other LSDBs are not affected

IR (Internal Router) definition: all interfaces are entirely within the same Area (all in Area 0 or all in Area 1), without crossing areas
ABR (Area Border Router) definition: one side is Area 0 and the other is a normal area; generates Type 3 summary LSAs; responsible for inter-area routes
ASBR (Autonomous System Boundary Router) definition: when an ABR device introduces external routes such as static routes, RIP, BGP, it becomes an ASBR; generates Type 5 LSAs (NSSA generates Type 7, converted to Type 5 when leaving the area)

Special Areas

Normal standard area: Types 1, 2, 3, 4, 5 LSAs all exist
Special areas are used to block external Types 4 and 5 LSAs; the ABR automatically advertises a default route to replace the huge number of specific routes

Reference detail table:

+++ info\