CloudEngine 8800, 7800, 6800, and 5800 Series

Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

1 Ethernet Switching

This chapter describes how to configure Ethernet switching.

1.1 Overview of Ethernet Switching

1.2 Basic Concepts of Ethernet
1.3 Switching on Ethernet
1.4 Application Scenarios for Ethernet Switching
1.5 Terms and Abbreviations

1.1 Overview of Ethernet Switching

Definition
The earliest Ethernet standard was the DEC-Intel-Xerox (DIX) standard jointly
developed by the Digital Equipment Corporation (DEC), Intel, and Xerox in 1982.
After years of development, Ethernet has become the most widely used local area
network (LAN) type, and many Ethernet standards have been put into use,
including standard Ethernet (10 Mbit/s), fast Ethernet (100 Mbit/s), gigabit
Ethernet (1000 Mbit/s), and 10G Ethernet (10 Gbit/s). IEEE 802.3 was defined
based on Ethernet and is compatible with Ethernet standards.
In the TCP/IP suite, the IP packet encapsulation format on an Ethernet network is
defined in RFC 894, and the IP packet encapsulation format on an IEEE 802.3
network is defined in RFC 1042. Currently, the format defined in RFC 894 is most
commonly used. This format is called Ethernet_II or Ethernet DIX.

NOTE

To distinguish Ethernet frames of the two types, Ethernet frames defined in RFC 894 are
called Ethernet_II frames and Ethernet frames defined in RFC 1042 IEEE 802.3 are called
frames in this document.

History
In 1972, when Robert Metcalfe (father of Ethernet) was hired by Xerox, his first
job was to connect computers in Xerox's Palo Alto Research Center (PARC) to the

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 1

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Advanced Research Projects Agency Network (ARPANET), progenitor of the

Internet. In 1972 also, Robert Metcalfe designed a network to connect computers
in the PARC. That network was based on the Aloha system (a radio network
system) and connected many computers in the PARC, so Metcalfe originally
named the network Alto Aloha network. The Alto Aloha network started
operating in May 1973, and Metcalfe then gave it an official name Ethernet, which
is the prototype of Ethernet. The network operated at a rate of 2.94 Mbit/s and
used thick coaxial cable as transmission medium. In June 1976, Metcalfe and his
assistant David Boggs published a paper Ethernet Distributed Packet Switching for
Local Computer Networks. At the end of 1977, Metcalfe and his three co-workers
were gained a patent on "Multipoint data communication system with collision
detection." Since then, Ethernet was known to the public.

As Ethernet technology develops rapidly, Ethernet has become the most widely
used LAN technology and replaced most of other LAN standards, such as token
ring, fiber distributed data interface (FDDI), and attached resource computer
network (ARCNET). After rapid development of 100M Ethernet in the 20th
century, gigabit Ethernet and even 10G Ethernet are now expanding their
applications as promoted by international standardization organizations and
industry-leading enterprises.

Purpose
Ethernet is a universal communication protocol standard used for local area
networks (LANs). This standard defines the cable type and signal processing
method used for LANs.

Ethernet networks are broadcast networks established based on the Carrier Sense
Multiple Access with Collision Detection (CSMA/CD) mechanism. Collisions restrict
Ethernet performance. Early Ethernet devices such as hubs work at the physical
layer, and cannot confine collisions to a particular scope. This restricts network
performance improvement. Working at the data link layer, switches are able to
confine collisions to a particular scope. Switches help improve Ethernet
performance and have replaced hubs as mainstream Ethernet devices. However,
switches do not restrict broadcast traffic on the Ethernet. This affects Ethernet
performance. Dividing a LAN into virtual local area networks (VLANs) on switches
or using Layer 3 switches can solve this problem.

As a simple, cost-effective, and easy-to-implement LAN technology, Ethernet has

become the mainstream in the industry. Gigabit Ethernet and even 10G Ethernet
make Ethernet the most promising network technology.

1.2 Basic Concepts of Ethernet

1.2.1 Ethernet Network Layers

Ethernet uses passive medium and transmits data in broadcast mode. It defines
protocols used on the physical layer and data link layer, interfaces between the
two layers, and interfaces between the data link layer and upper layers.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 2

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Physical Layer
The physical layer determines basic physical attributes of Ethernet, including data
coding, time scale, and electrical frequency.
The physical layer is the lowest layer in the Open Systems Interconnection (OSI)
reference model and is closest to the physical medium (communication channel)
that transmits data. Data is transmitted on the physical layer in binary bits (0 or
1). Transmission of bits depends on transmission devices and physical media, but
the physical layer does not refer to a specific physical device or a physical media.
Actually, the physical layer is located above a physical medium and provides the
data link layer with physical connections to transmit original bit streams.

Data Link Layer

The data link layer is the second layer in the OSI reference model, located
between the physical layer and network layer. The data link layer obtains service
from the physical layer and provides service for the network layer. The basic
service that the data link layer provides is to reliably transmit data from the
network layer of a source device to the network layer of an adjacent destination
device.
The physical layer and data link layer depend on each other. Therefore, different
working modes of the physical layer must be supported by corresponding data
link layer modes. This hinders Ethernet design and application.
Some organizations and vendors propose to divide the data link layer into two
sub-layers: the Media Access Control (MAC) sub-layer and the Logical Link Control
(LLC) sub-layer. Then different physical layers correspond to different MAC sub-
layers, and the LLC sub-layer becomes totally independent, as shown in Figure
1-1.

Figure 1-1 Hierarchy of Ethernet data link layer

Network
layer
LLC layer
Data link
layer MA
Physica layer
C
l
layer

The following sections describe concepts involved in the physical layer and data
link layer.

1.2.2 Introduction to Ethernet Cable Standards

Introduction to Ethernet Cable Standards

Currently, mature Ethernet physical layer standards are:
● 10BASE-2

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 3

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

● 10BASE-5
● 10BASE-T
● 10BASE-F
● 100BASE-T4
● 100BASE-TX
● 100BASE-FX
● 1000BASE-SX
● 1000BASE-LX
● 1000BASE-TX
● 10GBASE-T
● 10GBASE-LR
● 10GBASE-SR
In the preceding standards, 10, 100, 1000 and 10G stand for transmission rates,
and BASE represents baseband.
● 10M Ethernet cable standards
Table 1-1 lists the 10M Ethernet cable standards defined in IEEE 802.3.

Table 1-1 10M Ethernet cable standards

Name Cable Maximum
Transmission Distance

10BASE-5 Thick coaxial cable 500 m

10BASE-2 Thin coaxial cable 200 m

10BASE-T Twisted pair cable 100 m

10BASE-F Fiber 2000 m

NOTE

Coaxial cables have a fatal defect: Devices are connected in series and therefore a
single-point failure can cause the breakdown of the entire network. As the physical
standards of coaxial cables, 10BASE-2 and 10BASE-5 have fallen into disuse.
● 100M Ethernet cable standards
100M Ethernet is also called Fast Ethernet (FE). Compared with 10M Ethernet,
100M Ethernet has a faster transmission rate at the physical layer, but they
have no difference at the data link layer.
Table 1-2 lists the 100M Ethernet cable standards.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 4

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Table 1-2 100M Ethernet cable standards

Name Cable Maximum
Transmission Distance

100Base-T4 Four pairs of Category 3 100 m

twisted pair cables

100Base-TX Two pairs of Category 5 100 m

twisted pair cables

100Base-FX Single-mode fiber or multi- 2000 m

mode fiber

Both 10Base-T and 100Base-TX apply to Category 5 twisted pair cables. They
have different transmission rates. The 10Base-T transmits data at 10 Mbit/s,
whereas the 100Base-TX transmits data at 100 Mbit/s.
The 100Base-T4 is rarely used now.
● Gigabit Ethernet cable standards
Gigabit Ethernet is developed on the basis of the Ethernet standard defined in
IEEE 802.3. Based on the Ethernet protocol, Gigabit Ethernet increases the
transmission rate to 10 times the FE transmission rate, reaching 1 Gbit/s.
Table 1-3 lists the Gigabit Ethernet cable standards.

Table 1-3 Gigabit Ethernet cable standards

Interface Name Cables Maximum
Transmission Distance

1000Base-LX Single-mode fiber or multi- 316 m

mode fiber

1000Base-SX Multi-mode fiber 316 m

1000Base-TX Category 5 twisted pair 100 m

cable

Gigabit Ethernet technology can upgrade the existing Fast Ethernet from 100
Mbit/s to 1000 Mbit/s.
The physical layer of Gigabit Ethernet uses 8B10B coding. In traditional
Ethernet technology, the data link layer delivers 8-bit data sets to its physical
layer. After processing the data sets, the physical layer sends them to the data
link layer. The data sets are still 8 bits after processing.
The situation is different on the Gigabit Ethernet of optical fibers. The
physical layer maps the 8-bit data sets transmitted from the data link layer to
10-bit data sets and then sends them out.
● 10G Ethernet cable standards

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 5

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

10G Ethernet is currently defined in supplementary standard IEEE 802.3ae,

which will be combined with IEEE 802.3 later. Table 1-4 lists the 10G Ethernet
cable standards.

Table 1-4 10G Ethernet cable standards

Name Cables Maximum

Transmission Distance

10GBASE-T CAT-6A or CAT-7 100 m

10GBase-LR Single-mode optical fiber 10 km

10GBase-SR Multi-mode optical fiber Several hundred meters

● 100G Ethernet cable standards

The standard for 40G/100G Ethernet is defined in IEEE 802.3ba, which was
published in 2010. 100G Ethernet will be widely used as network technologies
develop.

1.2.3 CSMA/CD
● Definition of CSMA/CD
Ethernet was originally designed to connect computers and other digital
devices on a shared physical line. The computers and digital devices can
access the shared line only in half-duplex mode. Therefore, a mechanism of
collision detection and avoidance is required to prevent multiple devices from
contending for the line. This mechanism is called the carrier Sense Multiple
Access with Collision Detection (CSMA/CD).
The concept of CSMA/CD is described as follows:
– Carrier sense (CS)
Before transmitting data, a station checks whether the line is idle to
reduce chances of collision.
– Multiple access (MA)
Data sent by a station can be received by multiple stations.
– Collision detection (CD)
If two stations transmit electrical signals at the same time, the voltage
amplitude doubles the normal amplitude as signals of the two stations
accumulate. The situation results in collision.
The stations stop transmission after detecting the collision, and resume
the transmission after a random delay.
● CSMA/CD working process
CSMA/CD works as follows:
a. A station continuously detects whether the shared line is idle.

▪ If the line is idle, the station sends data.

▪ If the line is in use, the station waits until the line becomes idle.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 6

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

b. If two stations send data at the same time, a collision occurs on the line,
and signals on the line become unstable.
c. After detecting the instability, the station immediately stops sending
data.
d. The station sends a series of disturbing pulses. After a period of time, the
station resumes the data transmission.
The station sends disturbing pulses to inform other stations, especially
the station that sends data at the same time, that a collision occurred on
the line.
After detecting a collision, the station waits for a random period of time,
and then resumes the data transmission.

1.2.4 Minimum Frame Length and Maximum Transmission

Distance
Due to the limitation of the CSMA/CD algorithm, an Ethernet frame must be
longer than or equal to a specified length. On the Ethernet, the minimum frame
length is 64 bytes, which is determined jointly by the maximum transmission
distance and the collision detection mechanism.

The use of minimum frame length can prevent the following situation: station A
finishes sending the last bit, but the first bit does not arrive at station B, which is
far from station A. Station B considers that the line is idle and begins to send data,
leading to a collision.

Figure 1-2 Ethernet_II frame format

6bytes 6bytes 2bytes 46~1500bytes 4bytes

DMAC SMAC Type Data CRC

The upper layer protocol must ensure that the Data field of a packet contains at
least 46 bytes, so that the total length of the Data field, the 14-byte Ethernet
frame header, and the 4-byte check code at the frame tail can reach the minimum
frame length, as shown in Figure 1-2. If the Data field is less than 46 bytes, the
upper layer must pad the field to 46 bytes.

1.2.5 Duplex Modes of Ethernet

The physical layer of Ethernet can work in either half-duplex or full-duplex mode.

● Half-duplex mode
The half-duplex mode has the following features:
– Data can only be sent or received at any time.
– The CSMA/CD mechanism is used.
– The maximum transmission distance is limited.
Hubs work in half-duplex mode.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 7

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

● Full-duplex mode
After Layer 2 switches replace hubs, the shared Ethernet changes to the
switched Ethernet, and the half-duplex mode is replaced by the full-duplex
mode. As a result, the transmission rate increases greatly, and the maximum
throughput doubles the transmission rate.
The full-duplex mode solves the problem of collisions and eliminates the need
for the CSMA/CD mechanism.
The full-duplex mode has the following features:
– Data can be sent and received at the same time.
– The maximum throughput doubles the transmission rate.
– This mode does not have the limitation on the transmission distance.
All network cards, Layer 2 devices (except hubs), and Layer 3 devices
produced support the full-duplex mode.
The following hardware components are required to realize the full-duplex
mode:
– Full-duplex network cards and chips
– Physical media with separate data transmission and receiving channels
– Point-to-point connection

1.2.6 Auto-Negotiation of Ethernet

● Purpose of auto-negotiation
The earlier Ethernet adopts the 10 Mbit/s half-duplex mode; therefore,
mechanisms such as CSMA/CD are required to guarantee system stability.
With development of technologies, the full-duplex mode and 100M Ethernet
emerge, which greatly improve the Ethernet performance. How to achieve the
compatibility between the earlier and new Ethernet networks becomes a new
problem.
The auto-negotiation technology is introduced to solve this problem. In auto-
negotiation, the devices on two ends of a link can choose the same operation
parameters by exchanging information. The main parameters to be
negotiated are mode (half-duplex or full-duplex), speed, and flow control.
After the negotiation succeeds, the devices on two ends operate in the
negotiated mode and rate.
The auto-negotiation of duplex mode and speed is defined in the following
standards:
– 100M Ethernet standard: IEEE 802.3u
In IEEE 802.3u, auto-negotiation is defined as an optional function.
– Gigabit Ethernet standard: IEEE 802.3z
In IEEE 802.3z, auto-negotiation is defined as a mandatory and default
function.
● Principle of auto-negotiation
Auto-negotiation is an Ethernet procedure by which two connected devices
choose common transmission parameters. It allows a network device to
transmit the supported operating mode to the peer and receives the operating
mode from the peer. In this process, the connected devices first share their

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 8

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

capabilities regarding these parameters and then choose the highest

performance transmission mode they both support.
When no data is transmitted over a twisted pair on an Ethernet network,
pulses of high frequency are transmitted at an interval of 16 ms to maintain
the connections at the link layer. These pulses form a Normal Link Pulse
(NLP) code stream. Some pulses of higher frequency can be inserted in the
NLP to transmit more information. These pulses form a Fast Link Pulse (FLP)
code stream, as shown in Figure 1-3. The basic mechanism of auto-
negotiation is to encapsulate the negotiation information into FLP.

Figure 1-3 Pulse insertion

m
16
s
1ms

16 small pulses are inserted

into every pulse

Similar to an Ethernet network that uses twisted pair cables, an Ethernet

network that uses optical modules and optical fibers also implements auto-
negotiation by sending code streams. These code streams are called
Configuration (C) code streams. Different from electrical interfaces, optical
interfaces do not negotiate traffic transmission rates and they work in duplex
mode. Optical interfaces only negotiate flow control parameters.
If auto-negotiation succeeds, the Ethernet card activates the link. Then, data
can be transmitted on the link. If auto-negotiation fails, the link is
unavailable.
If one end does not support auto-negotiation, the other end that supports
auto-negotiation adopts the default operating mode, which is generally 10
Mbit/s half-duplex.
Auto-negotiation is implemented based on the chip design at the physical
layer. As defined in IEEE 802.3, auto-negotiation is implemented in any of the
following cases:
– A faulty link recovers.
– A device is power recycled.
– Either of two connected devices resets.
– A renegotiation request packet is received.
In other cases, two connected devices do not always send auto-negotiation
code streams. Auto-negotiation does not use special packets or bring
additional protocol costs.
● Auto-negotiation rules for interfaces
Two connected interfaces can communicate with each other only when they
are working in the same working mode.
– If both interfaces work in the same non-auto-negotiation mode, the
interfaces can communicate.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 9

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

– If both interfaces work in auto-negotiation mode, the interfaces can

communicate through negotiation. The negotiated working mode
depends on the interface with lower capability (specifically, if one
interface works in full-duplex mode and the other interface works in half-
duplex mode, the negotiated working mode is half-duplex). The auto-
negotiation function also allows the interfaces to negotiate about the
flow control function.
– If a local interface works in auto-negotiation mode and the remote
interface works in a non-auto-negotiation mode, the negotiated working
mode of the local interface depends on the working mode of the remote
interface.

1.2.7 Collision Domain and Broadcast Domain

Collision Domain
On a legacy Ethernet network using thick coaxial cables as a transmission
medium, multiple nodes on a shared medium share the bandwidth on the link and
compete for the right to use the link. A network collision occurs when more than
one node attempts to send a packet on this link at the same time. The carrier
sense multiple access with collision detection (CSMA/CD) mechanism is used to
solve the problem of collisions. Once a collision occurs on a link, the CSMA/CD
mechanism prevents data transmission on this link within a specified time.
Collisions are inevitable on an Ethernet network, and the probability that collision
occurs increases when more nodes are deployed on a shared medium. All nodes
on a shared medium constitute a collision domain. All the nodes in a collision
domain compete for bandwidth. Packets sent from a node, including unicast,
multicast, and broadcast packets, can reach all the other nodes in the collision
domain.

Broadcast Domain
Packets are broadcast in a collision domain, which results in a low bandwidth
efficiency and degrades packet processing performance of network devices.
Therefore, broadcasting of packets must be restricted. For example, the ARP
protocol sends broadcast packets to obtain MAC addresses mapping specified IP
addresses. The all 1s MAC address FFFF-FFFF-FFFF is the broadcast MAC address.
All nodes must process data frames with this MAC address as the destination MAC
address. A broadcast domain is a group of nodes, among which broadcast packet
from one node can reach all the other nodes. A network bridge forwards unicast
packets according to its MAC address table and forwards broadcast packets to all
its ports. Therefore, nodes connected to all ports of a bridge belong to a broadcast
domain, but each port belongs to a different collision domain.

1.2.8 MAC Sub-layer

Functions of the MAC Sub-layer
The MAC sub-layer has the following functions:
● Provides access to physical links.
The MAC sub-layer is associated with the physical layer. That is, different MAC
sub-layers provide access to different physical layers.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 10

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Ethernet has two types of MAC sub-layers:

– Half-duplex MAC: provides access to the physical layer in half-duplex
mode.
– Full-duplex MAC: provides access to the physical layer in full-duplex
mode.
The two types of MAC sub-layers are integrated in a network interface card.
After the network interface card is initialized, auto-negotiation is performed
to choose an operation mode, and then a MAC sub-layer is chosen according
to the operation mode.
● Identifies stations at the data link layer.
The MAC sub-layer reserves a unique MAC address for each station.
The MAC sub-layer uses a MAC address to uniquely identify a station.
MAC addresses are managed by Institute of Electrical and Electronics
Engineers (IEEE) and allocated in blocks. An organization, generally a device
manufacturer, obtains a unique address block from IEEE. The address block is
called an Organizationally Unique Identifier (OUI). Using the OUI, the
organization can allocate MAC addresses to 16777216 devices.
A MAC address has 48 bits, which are generally expressed in 12-digit dotted
hexadecimal notation. For example, the 48-bit MAC address
000000001110000011111100001110011000000000110100 is represented by
00e0.fc39.8034.
The first 6 digits in dotted hexadecimal notation stand for the OUI, and the
last 6 digits are allocated by the vendor. For example, in 00e0.fc39.8034,
00e0.fc is the OUI allocated by IEEE to Huawei, and 39.8034 is the address
number allocated by Huawei.
The second bit of a MAC address indicates whether the address is globally
unique or locally unique. Ethernet uses globally unique MAC addresses.
MAC addresses are divided into the following types:
– Physical MAC address
A physical MAC address is burned into hardware (such as a network
interface card) and uniquely identifies a terminal on the Ethernet.
– Broadcast MAC address
A broadcast MAC address indicates all the terminals on a network.
The 48 bits of a broadcast MAC address are all 1s, such as ffff.ffff.ffff.
– Multicast MAC address
A multicast MAC address indicates a group of terminals on a network.
The eighth bit of a multicast MAC address is 1, such as
000000011011101100111010101110101011111010101000.
● Transmits data over the data link layer. After receiving data from the LLC sub-
layer, the MAC sub-layer adds the MAC address and control information to
the data, and then transmits the data to the physical link. In the process, the
MAC sub-layer provides other functions such as the check function.
Data is transmitted at the data link layer as follows:
a. The upper layer delivers data to the MAC sub-layer.
b. The MAC sub-layer stores the data in the buffer.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 11

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

c. The MAC sub-layer adds the destination MAC address and source MAC
address to the data, calculates the length of the data frame, and forms
an Ethernet frame.
d. The Ethernet frame is sent to the peer according to the destination MAC
address.
e. The peer compares the destination MAC address with entries in the MAC
address table.

▪ If a matching entry is found, the frame is accepted.

▪ If no matching entry is found, the frame is discarded.

The preceding describes frame transmission in unicast mode. After an upper-
layer application is added to a multicast group, the data link layer generates a
multicast MAC address according to the application, and then adds the
multicast MAC address to the MAC address table. The MAC sub-layer receives
frames with the multicast MAC address and transmits the frames to the upper
layer.

Ethernet Frame Structure

● Format of an Ethernet_II frame

Figure 1-4 Format of an Ethernet_II frame

6bytes 6bytes 2bytes 46~1500bytes 4bytes
DMAC SMAC Type Data CRC

Table 1-5 describes the fields in an Ethernet_II frame.

Table 1-5 Fields in an Ethernet_II frame

Field Description

DMAC It indicates the destination MAC address. DMAC specifies

the receiver of the frame.

SMAC It indicates the source MAC address. SMAC specifies the

station that sends the frame.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 12

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Field Description

Type The 2-byte Type field identifies the upper layer protocol of
the Data field. The receiver can know the meaning of the
Data field according to the Type field.
Ethernet allows multiple protocols to coexist on a LAN. The
hexadecimal values in the Type field of an Ethernet_II frame
stand for different protocols.
● Frames with the Type field value 0800 are IP frames.
● Frames with the Type field value 0806 are Address
Resolution Protocol (ARP) frames.
● Frames with the Type field value 8035 are Reverse
Address Resolution Protocol (RARP) frames.
● Frames with the Type field value 8137 are Internetwork
Packet Exchange (IPx) and Sequenced Packet Exchange
(SPx) frames.

Data The minimum length of the Data field is 46 bytes, which

ensures that the frame is at least 64 bytes in length. The
46-byte Data field is required even if only 1-byte
information needs to be transmitted.
If the payload of the Data field is less than 46 bytes, the
Data field must be padded to 46 bytes.
The maximum length of the Data field is 1500 bytes.

CRC The Cyclic Redundancy Check (CRC) field provides an error

detection mechanism.
Each sending device calculates a CRC code containing the
DMAC, SMAC, Type, and Data fields. Then the CRC code is
filled into the 4-byte CRC field.

● Format of an IEEE 802.3 frame

Figure 1-5 Format of an IEEE 802.3 frame

6byte 6byte 2byte 38~1492byte 4byte
DMAC SMAC Length LLC SNAP Data CRC

DSAP SSAPControl org code Type

1byte 1byte 1byte 3byte 2byte

As shown in Figure 1-5, the format of an IEEE 802.3 frame is similar to that
of an Ethernet_II frame except that the Type field is changed to the Length
field in an IEEE 802.3 frame, and the LLC field and the Sub-Network Access
Protocol (SNAP) field occupy 8 bytes of the Data field.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 13

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Table 1-6 Format of an IEEE 802.3 frame

Field Description

Length The Length field specifies the number of bytes in the Data
field.

LLC The LLC field consists of three sub-fields: Destination Service

Access Point (DSAP), Source Service Access Point (SSAP),
and Control.

SNAP The SNAP field consists of the Org Code field and the Type
field. Three bytes in the Org Code field are all 0s. The Type
field functions the same as the Type field in Ethernet_II
frames.

NOTE

For description about other fields, see the description of Ethernet_II frames.
Based on the values of DSAP and SSAP, IEEE 802.3 frames can be divided into
the following types:
– If DSAP and SSAP are both 0xff, the IEEE 802.3 frame changes to a
Netware-Ethernet frame that carries NetWare data.
– If DSAP and SSAP are both 0xaa, the IEEE 802.3 frame changes to an
Ethernet_SNAP frame.
Ethernet_SNAP frames can be encapsulated with data of multiple
protocols. The SNAP can be considered as an extension of the Ethernet
protocol. SNAP allows vendors to define their own Ethernet transmission
protocols.
The Ethernet_SNAP standard is defined by IEEE 802.1 to guarantee
interoperability between IEEE 802.3 LANs and Ethernet networks.
– Other values of DSAP and SSAP indicate IEEE 802.3 frames.

1.2.9 LLC Sub-layer

The MAC sub-layer supports two types of frame: IEEE 802.3 frames and Ethernet_II
frames. In an Ethernet_II frame, the Type field identifies the upper layer protocol.
Therefore, only the MAC sub-layer is required on a device, and the LLC sub-layer
does not need to be realized.
In an IEEE 802.3 frame, the LLC sub-layer defines useful features in addition to
traditional services of the data link layer. All these features are provided by the
sub-fields of DSAP, SSAP, and Control.
The following lists three types of point-to-point services:
● Connectionless service
Currently, the Ethernet implements this service.
● Connection-oriented service
A connection is set up before data is transmitted. The reliability of data is
guaranteed during the transmission.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 14

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

● Connectionless data transmission with acknowledgement

A connection is not required before data transmission. The acknowledgement
mechanism is used to improve the reliability.

The following is an example that describes the applications of SSAP and DSAP.
Assume that terminals A and B use connection-oriented services. Data is
transmitted in the following process:

1. A sends a frame to B to require the establishment of a connection with B.

2. If B has enough resources, it returns an acknowledgement message that
contains a Service Access Point (SAP). The SAP identifies the connection
required by A.
3. After receiving the acknowledgement message, A knows that B has set up a
local connection with A. After creating an SAP, A sends a message containing
the SAP to B. The connection is set up.
4. The LLC sub-layer of A encapsulates the data into a frame. The DSAP field is
filled in with the SAP sent by B; the SSAP field is filled in with the SAP created
by A. Then the LLC sub-layer sends the frame to the MAC sub-layer of A.
5. The MAC sub-layer of A adds the MAC address and the Length field into the
frame, and then sends the frame to the data link layer.
6. After the frame is received at the MAC sub-layer of B, the frame is
transmitted to the LLC sub-layer. The LLC sub-layer figures out the connection
to which the frame belongs according to the DSAP field.
7. After checking and acknowledging the frame based on the connection type,
the LLC sub-layer of B transmits the frame to the upper layer.
8. After the frame reaches its destination, A instructs B to release the connection
by sending a frame. At this time, the communications end.

1.3 Switching on Ethernet

1.3.1 Layer 2 Switching

A Layer 2 device works at the second layer of the OSI model and forwards data
packets based on media access control (MAC) addresses. Ports on a Layer 2 device
send and receive data independently and belong to different collision domains.
Collision domains are isolated at the physical layer so that collisions will not occur
between hosts (or networks) connected through this Layer 2 device due to uneven
traffic rates on these hosts (or networks).

A Layer 2 device parses and learns source MAC addresses of Ethernet frames and
maintains a mapping table of MAC addresses and ports. This table is called a MAC
address table. When receiving an Ethernet frame, the device searches for the
destination MAC address of the frame in the MAC table to determine through
which port to forward this frame.

1. When the Layer 2 device receives an Ethernet frame, it records the source
MAC address and the inbound port of the frame in the MAC address table to
guide Layer 2 forwarding. If the same MAC address entry exists in the MAC
address table, the device resets the aging time of the entry. An aging
mechanism is used to maintain entries in the MAC address table. Entries that

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 15

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

are not updated within the aging time are deleted from the MAC address
table.
2. The device looks up the MAC address table based on the destination MAC
address of the Ethernet frame. If no matching entry is found, the device
forwards the frame to all its ports except the port from which the frame is
received. If the destination MAC address of the frame is a broadcast address,
the device forwards the frame to all its ports except the port from which the
frame is received. If a matching entry is found in the MAC address table, the
device forwards the frame to the port specified in the entry.

According to the preceding forwarding process, a Layer 2 device maintains a MAC

address table and forwards Ethernet frames based on destination MAC addresses.
This forwarding mechanism fully uses network bandwidth and improves network
performance. Figure 1-6 shows an example of Layer 2 switching

Figure 1-6 Layer 2 switching example

MAC Address Port
MAC A Port 1
PC B
MAC B Port 2
MAC C Port 3

PC A Port 2
Port 1
Port 3
PC C
MAC C MAC A Type Data MA
CC
MA
CA
Typ
e Dat
a

Although Layer 2 devices can isolate collision domains, they cannot isolate
broadcast domains. As described in the Layer 2 forwarding process, broadcast
packets and packets that do not match nay entry in the MAC address table are
forwarded to all ports (except the port from which the frame is received). Packet
broadcasting consumes much bandwidth on network links and brings security
issues. Routers can isolate broadcast domains, but high costs and low forwarding
performance of routers limit the application of routers in Layer 2 forwarding. The
virtual local area network (VLAN) technology is introduced to solve this problem
in Layer 2 switching.

1.3.2 Layer 3 Switching

Background of Layer 3 Switches

In early stage of network deployment, most local area networks (LANs) were
established using Layer 2 switches, and routers completed communication
between LANs. At that time, intra-LAN traffic accounted for most of network
traffic and little traffic was transmitted between LANs. A few routers were enough
to handle traffic transmission between LANs.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 16

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

As data communication networks expand and more services emerge on the

networks, increasing traffic needs to be transmitted between networks. Routers
cannot adapt to this development trend because of their high costs, low
forwarding performance, and small port quantities. New devices capable of high-
speed Layer 3 forwarding are required. Layer 3 switches are such devices.
Routers use CPUs to complete Layer 3 forwarding, whereas Layer 3 switches use
hardware to complete Layer 3 forwarding. Hardware forwarding has a much
higher performance than software forwarding (CPU based forwarding). Switches
cannot replace routers in all scenarios because routers provide rich interface types,
good service class control, and powerful routing capabilities that Layer 3 switches
cannot provide.

Layer 3 Forwarding Mechanism

Layer 3 switches divide a Layer 2 network into multiple VLANs. They implement
Layer 2 switching within the VLANs and Layer 3 IP connectivity between VLANs.
Two hosts on different networks communicate with each other through the
following process:
1. Before the source host starts communicating with the destination host, it
compares its own IP address with the IP address of the destination host. If IP
addresses of the two hosts have the same network ID (calculated by an AND
operation between the IP addresses and masks), the hosts are located on the
same network segment. In this case, the source host sends an Address
Resolution Protocol (ARP) request to the destination host. After receiving an
ARP reply from the destination host, the source host obtains the MAC address
of the destination host and sends packets to this destination MAC address.
2. If the source and destination hosts are located on different network segments,
the source host sends an ARP request to obtain the MAC address mapping the
gateway IP address. After receiving an ARP reply from the gateway, the source
host sends packets to the MAC address of the gateway. In these packets, the
source IP address is the IP address of the source host, and destination IP
address is still the IP address of the destination host.
The following is the detailed Layer 3 switching process.
As shown in Figure 1-7, the source and destination hosts connect to the same
Layer 3 switch but belong to different VLANs (network segments). Both the two
hosts are located on the directly connected network segments of the Layer 3
switch, so the routes to the IP addresses of the hosts are direct routes.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 17

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Figure 1-7 Layer 3 forwarding

PC A PC B
MAC A MAC B
IP:10.1.1.2 L3 Switch IP:10.2.1.2
GW:10.1.1.1 MAC Switch GW:10.2.1.1
VLAN 2 10.1.1.1 10.2.1.1 VLAN 3

DMAC:MAC Switch DMAC:MAC B

SMAC:MAC A SMAC:MAC Switch
DIP: 10.2.1.2 DIP: 10.2.1.2
SIP: 10.1.1.2 SIP: 10.1.1.2

Figure 1-7 shows the MAC addresses, IP addresses, and gateway addresses of the
hosts, MAC address of the Layer 3 switch, and IP addresses of Layer 3 interfaces
configured in VLANs on the Layer 3 switch. The process of a ping from PC A to PC
B is as follows (the Layer 3 switch has not created any MAC address entry):
1. PC A finds that the destination IP address 10.2.1.2 (PC B) is on a different
network segment than its own IP address. Therefore, PC A sends an ARP
request to request for the MAC address mapping the gateway address
10.1.1.1.
2. L3 Switch receives the ARP request from PC A and finds that 10.1.1.1 is the IP
address of its own Layer 3 interface. L3 switch then sends an ARP reply to PC
A. The ARP reply carries the MAC address of its Layer 3 interface (MAC
Switch). In addition, L3 switch adds the mapping between the IP address and
MAC address of PC A (10.1.1.2 and MAC A) to its ARP table. The IP address
and MAC address of PC A are carried in the ARP request sent from PC A. And
L3 Switch adds the mapping between the source MAC address and VLAN ID
of the packet and outbound port to its MAC table.
3. After PC A receives the ARP reply from the gateway (L3 Switch), it sends an
ICMP request packet. In the ICMP request packet, the destination MAC
address (DMAC) is MAC Switch; the source MAC address (SMAC) is MAC A;
the source IP address (SIP) is 10.1.1.2; the destination IP address (DIP) is
10.2.1.2.
4. When L3 Switch receives the ICMP request packet, it looks up the MAC
address table according to the destination MAC address and VLAN ID of the
packet and finds the entry with the MAC address of its Layer 3 interface, the
packet needs to be forwarded at Layer 3. Then L3 Switch looks up Layer 3
forwarding entries of the switching chip to guide Layer 3 forwarding.
5. The switching chip loops up Layer 3 forwarding entries according to the
destination IP address of the packet. The entry lookup fails because no entry
has been created. The switching chip then sends the packet to the CPU for
software processing.
6. The CPU looks up the software routing table according to the destination IP
address of the packet and finds a directly connected network segment,
network segment of PC B. Then the CPU looks up its ARP table, and the
lookup still fails. Therefore, L3 Switch sends an ARP request to all ports in
VLAN 3 (network segment of PC B), to request the MAC address mapping IP
address 10.2.1.2.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 18

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

7. After PC B receives the ARP request from L3 Switch, it checks the ARP request
and finds that 10.2.1.2 is its own IP address. PC B then sends an ARP reply
carrying its MAC address (MAC B). Meanwhile, PC B records the mapping
between the IP address and MAC address of L3 Switch (10.2.1.1 and MAC
Switch) in its ARP table.
8. When L3 Switch receives the ARP reply from PC B, it records the mapping
between the IP address and MAC address of PC B (10.2.1.2 and MAC B) in its
ARP table. L3 Switch changes the destination MAC address in the ICMP
request packet sent from PC A to MAC B and changes the source MAC
address to its own MAC address (MAC Switch), and then sends the ICMP
request to PC B. The Layer 3 forwarding entry containing the IP address and
MAC address of PC B, outbound VLAN ID, and outbound port is also added to
the Layer 3 forwarding of the switching chip. Subsequent packets sent from
PC A to PC B are directly forwarded according to this hardware entry.
9. When PC B receives the ICMP request packet from L3 Switch, it sends an ICMP
reply packet to PC A. The forwarding process for the ICMP reply packet is
similar to that for the ICMP request packet except that the ICMP reply packet
is directly forwarded to PC A by the switching chip according to the hardware
entry. The reason is that L3 Switch has obtained the mapping between the IP
address and MAC address of PC A and added matching Layer 3 forwarding
entry to the L3 forwarding table of the switching chip.
10. Subsequent packets exchanged between PC A and PC B are forwarded
following the same process: MAC address table lookup, Layer 3 forwarding
table lookup, and hardware forwarding by the switching chip.
In a summary, a Layer 3 switch provides high-speed Layer 3 switching through one
routing process (forwarding the first packet to the CPU and creating a hardware
Layer 3 forwarding entry) and multiple switching processes (hardware forwarding
of subsequent packets).

1.4 Application Scenarios for Ethernet Switching

1.4.1 Building a Data Center Network

A data center has a series of complex facilities. Besides a computer system and
cooperating devices (such as communications devices and storage systems), a
data center requires redundant data communication connections, environment
control devices, monitoring devices, and security devices.
Figure 1-8 shows a typical data center network topology. On this network,
switches connect to each other and to servers using Ethernet technology and have
various services deployed to provide low-latency, high-reliability services under
highly concurrent scenarios.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 19

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Figure 1-8 Using Ethernet technology to build a data center network

Data center
backbone
network

Core Layer

Access Layer

Server

As services in a data center require high network performance, 10G or even

higher-speed Ethernet has been widely used in data centers. Data centers create
chances for use of advanced Ethernet technologies and promote development of
Ethernet technologies.

1.5 Terms and Abbreviations

Terms
Term Description

10Base-T Defined in IEEE 802.3i, it is an Ethernet specification that

uses the twist pair with the maximum length of 100 meters
(328.08 ft.) at 10 Mbit/s for each network segment.

100Base-T Defined in IEEE 802.3u, it is a Fast Ethernet specification that

uses the twist pair with the maximum length of 100 meters
(328.08 ft.) at 100 Mbit/s for each network segment.

1000BaseT Defined in IEEE 802.3ab, it is an Ethernet specification that

uses the twist pair with the maximum length of 100 meters
(328.08 ft.) at 1000 Mbit/s for each network segment.

Ethernet Created by Xerox and developed by Xerox, Intel, and Digital

Equipment Corporation (DEC), it is a baseband LAN
specification that uses CSMA/CD and transmits data over
various cables at 10 Mbit/s. Ethernet-related standards are
defined in IEEE 802.3 series.

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 20

CloudEngine 8800, 7800, 6800, and 5800 Series
Switches
Configuration Guide - Ethernet Switching 1 Ethernet Switching

Term Description

Ethernet_II An encapsulation format of Ethernet frames, which is the

standard ARPA Ethernet Version 2.0 encapsulation that uses
a 16-bit protocol type code.

Ethernet_SNAP An encapsulation format of Ethernet frames. As specified in

RFC 1042, it allows Ethernet frames to be transmitted
through IEEE 802.2 media.

FE Short for the Fast Ethernet. Complying with IEEE 802.3u, it is

an extension and enhancement of the traditional media-
sharing Ethernet standard and allows data to be transmitted
at 100 Mbit/s.

Full-duplex The full-duplex mode is an operation mode of Ethernet

interfaces. In full-duplex mode, interfaces on both ends can
send and receive data at the same time without interruption.

GE Short for Gigabit Ethernet. Complying with IEEE 802.3z, the

GE is compatible with the 10M Ethernet and the 100M
Ethernet (FE).

Half-duplex An operation mode of Ethernet interfaces. In half-duplex

mode, an interface can only receive or send data at a time.

MAC Short for Media Access Control. At the data link layer of the
OSI model, the MAC sub-layer is adjacent to the physical
layer.

Auto- A function that enables devices on both ends of a physical

negotiation link to automatically select an operation mode by
exchanging information. In auto-negotiation, the duplex
mode and operation rate are negotiated. Once the
negotiation result is approved, the operation mode is fixed
until the device is restarted or the cable is removed.

Abbreviations
Abbreviation Full Name

CSMA/CD Carrier Sense Multiple Access with Collision Detection

GE Gigabit Ethernet

MAC Media Access Control

TCP Transmission Control Protocol

Issue 09 (2021-06-03) Copyright © Huawei Technologies Co., Ltd. 21