Professional Documents
Culture Documents
Issue 02
Date 2017-06-29
and other Huawei trademarks are trademarks of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their respective
holders.
Notice
The purchased products, services and features are stipulated by the contract made between Huawei and the
customer. All or part of the products, services and features described in this document may not be within the
purchase scope or the usage scope. Unless otherwise specified in the contract, all statements, information,
and recommendations in this document are provided "AS IS" without warranties, guarantees or
representations of any kind, either express or implied.
The information in this document is subject to change without notice. Every effort has been made in the
preparation of this document to ensure accuracy of the contents, but all statements, information, and
recommendations in this document do not constitute a warranty of any kind, express or implied.
Website: http://www.huawei.com
Email: support@huawei.com
Contents
2 Overview......................................................................................................................................... 4
3 Technical Description...................................................................................................................5
4 Related Features.............................................................................................................................7
5 Network Impact............................................................................................................................. 8
6 Engineering Guidelines............................................................................................................... 9
6.1 When to Use................................................................................................................................................................... 9
6.2 Required Information..................................................................................................................................................... 9
6.3 Planning........................................................................................................................................................................ 10
6.4 Deployment.................................................................................................................................................................. 10
6.4.1 Requirements............................................................................................................................................................. 10
6.4.2 Data Preparation and Feature Activation...................................................................................................................11
6.4.2.1 Data Preparation..................................................................................................................................................... 11
6.4.2.2 Activation................................................................................................................................................................11
6.4.3 Activation Observation.............................................................................................................................................. 11
6.4.4 Reconfiguration......................................................................................................................................................... 12
6.4.5 Deactivation...............................................................................................................................................................12
6.5 Performance Monitoring...............................................................................................................................................13
6.6 Parameter Optimization................................................................................................................................................ 14
6.7 Possible Issues.............................................................................................................................................................. 14
7 Parameters..................................................................................................................................... 15
8 Counters........................................................................................................................................ 17
9 Glossary......................................................................................................................................... 22
10 Reference Documents............................................................................................................... 23
1.1 Scope
This document describes LBFD-001027/MLBFD-12100242 Active Queue Management
(AQM), including its technical principles, related features, network impact, and engineering
guidelines.
Feature compatibility with specific terminal models is not presented in this document. For
compatibility information, contact Huawei engineers.
Any parameters, alarms, counters, or managed objects (MOs) described herein apply only to
the corresponding software release. For future software releases, refer to the corresponding
updated product documentation.
This document applies only to LTE FDD and LTE NB-IoT. Any "LTE" in this document
refers to LTE FDD or LTE NB-IoT, and "eNodeB" refers to LTE FDD eNodeB or LTE NB-
IoT eNodeB.
l Feature change
Changes in features and parameters of a specified version as well as the affected entities
l Editorial change
Changes in wording or addition of information and any related parameters affected by
editorial changes. Editorial change does not specify the affected entities.
eRAN12.1 02 (2017-06-29)
This issue includes the following changes.
eRAN12.1 01 (2017-03-08)
This issue does not include any changes.
2 Overview
A radio link usually has a lower bandwidth than a wired link and the bandwidth of the radio
link fluctuates more greatly than that of the wired link. In the downlink hybrid service
scenario where a bearer contains data of multiple Transmission Control Protocol (TCP)
services, some relatively prevailing data usually exists. The data occupies air interface
resources for a long time, resulting in resource insufficiency. As a result, a large amount of
data is buffered on the eNodeB, causing congestion. AQM, which proactively controls the
queue length, is introduced to minimize congestion.
When air interface resources are insufficient, AQM estimates the length of the packet queue
of bearer services and adaptively drops TCP packets in case of congestion. Congestion control
at the sender is triggered in advance to prevent congestion deterioration.
AQM offers the following benefits:
l Proactively detects possible congestion in the queues and minimizes the congestion
probability by dropping packets.
l Maintains shorter queues to balance the throughput and transmission delay.
l Avoids TCP global synchronization caused by simultaneous data overflow of multiple
queues.
NOTE
Data traffic is generally bursty, and data packets arrive at a queue suddenly. A large number of packets
will be dropped when the queue becomes full, or is getting close to becoming full. Upon packet loss
detection, multiple TCP senders adaptively and sharply decrease their transmission rates. Consequently,
the packet arrival rate declines quickly and the network congestion is relieved. After detecting that the
network is no longer congested, the TCP senders start to increase transmission rates, which again results
in network congestion. In the procedure, each sender decreases and increases transmission rates at the
same time as other senders. The phenomenon is called TCP global synchronization.
3 Technical Description
Congestion has become a bottleneck in network development due to the limitation of resource
capacity and processing capability of networks. End-to-end TCP congestion control
techniques have been the primary methods to control IP network congestion. However, edge
nodes in a network may not be able to obtain accurate information about network status in
time. Therefore, it is hard for these TCP congestion control techniques to solve all congestion
problems in the network. To address this issue, congestion control techniques based on
intermediate network nodes, such as AQM, have been developed.
As an IP network node, the eNodeB uses AQM to check packet queues of bearer services. If
AQM detects that the queue length (measured in packets) exceeds a specified threshold, it
drops packets selectively. This reduces the amount of data to be sent, mitigates the congestion,
reduces the delay of interactive services, and improves scheduling fairness among services.
As shown in Figure 3-1, AQM works in normal, congestion avoidance, and congestion areas.
These three areas are separated by minimum and maximum congestion thresholds, where the
red lines denote the packet loss rate and Max represents the maximum packet loss rate of a
non-GBR service in the congestion avoidance area.
AQM calculates the average queue length and compares it with the minimum and maximum
congestion thresholds. It then performs the following operations based on the comparison
result:
l Average queue length ≤ Minimum congestion threshold (normal area)
AQM does not drop any packets because the queue is not congested.
l Minimum congestion threshold < Average queue length < Maximum congestion
threshold (congestion avoidance area)
AQM drops some packets because the queue is congested to some extent. The
probability of dropping packets increases linearly with:
– The average queue length
– The amount of data successfully sent from the core network to the eNodeB
l Average queue length ≥ Maximum congestion threshold (congestion control area)
AQM drops every newly arrived packet because the queue is severely congested.
The minimum congestion threshold is automatically calculated by the eNodeB based on the
real-time air interface data rate. The maximum congestion threshold is equal to the minimum
congestion threshold multiplied by eight.
4 Related Features
Prerequisite Features
None
Impacted Features
None
5 Network Impact
System Capacity
When AQM is enabled, the cell throughput fluctuates as follows:
l If a cell has both cell edge users (CEUs) and cell center users (CCUs), AQM will
decrease the throughput of CEUs because CEU services are easily congested.
Alternatively, CCU services are rarely congested, and therefore AQM will increase the
throughput of CCUs. This increased throughput is larger than the decreased throughput
for CEUs. In this case, the overall cell throughput increases.
l If a cell has only CEUs or CCUs, AQM has little impact on cell throughput.
Network Performance
When AQM is enabled, the average processing delay of downlink packets carried by non-
GBR services decreases and therefore, the delay of delay-sensitive services (such as web
browsing and small file downloading) also decreases. In addition, the throughput is affected,
for example, the cell throughput and the throughput of some UEs fluctuate, and the
throughput for non-GBR services decreases. If GBR and non-GBR services coexist in a cell
and the actual GBR service rate is less than the subscribed GBR service rate, the decrease in
the throughput of non-GBR services possibly brings more scheduling opportunities to GBR
services, thereby increasing the throughput of GBR services.
6 Engineering Guidelines
l The packet loss rate in specific cells is abnormally high and user experience in
interactive services deteriorates because a large amount of data is present in the
downlink.
l A large number of TCP-based bit torrent (BT) services or large-sized file download
services are present on the network while users are using many interactive services.
In the following scenarios, AQM offers no gains or could even have slightly negative effects:
l AQM helps reduce only TCP traffic on a bearer when both TCP and User Datagram
Protocol (UDP) services are carried on the bearer.
l AQM drops packets randomly during network congestion. Therefore, if AQM drops
packets of interactive services, user experience in these services deteriorates.
l The eNodeB does not enter the congestion state when the data rates of CEUs fluctuate
due to interference, and therefore AQM offers no obvious gains for such UEs.
l Dropping packets causes throughput fluctuation, and the fluctuation is more obvious
when there are only a small number of TCP connections, for example, one or two
connections.
l Non-GBR services of a UE will not be congested when the downlink UE throughput on
a bearer exceeds 30 Mbit/s. In this scenario, AQM offers no obvious gains.
It is recommended that AQM be enabled if the average processing delay of downlink PDCP
SDUs for services carried on DRBs is larger than 500 ms per packet for a specific QCI and
the value of the Discard timer (ms) parameter for the QCI is larger than 500 ms. The Discard
timer (ms) parameter value can be queried by running the LST RLCPDCPPARAGROUP
command.
NOTE
The average processing delay of downlink PDCP SDUs for services carried on DRBs can be monitored
by dividing L.Traffic.DL.PktDelay.Time.QCI.x by L.Traffic.DL.PktDelay.Num.QCI.x, where x indicates
a QoS class identifier (QCI).
6.3 Planning
None
6.4 Deployment
6.4.1 Requirements
This feature has no requirements on related features, hardware, and licenses.
To deploy MLBFD-12100242 Active Queue Management (AQM), you need to learn NB-
IoT's requirements for hardware. For details, see NB-IoT Radio and Performance Basics
Feature Parameter Description.
6.4.2.2 Activation
The size of a small file can be determined based on the air interface throughput of the UE. If the
time for downloading a file while downloading one or multiple large files is within 10s, the file is
a small file.
Step 1 Run the MOD CELLALGOSWITCH command with the AqmAlgoSwitch option of the
AQM algorithm switch parameter deselected to disable AQM. Then, enable the UE to access
the network.
Step 2 Download one or more large files and a small file simultaneously. Then, record the small file
download latency. Repeat the test and average the recorded values. The average value works
as the small file download latency with AQM disabled.
Step 3 Run the MOD CELLALGOSWITCH command with the AqmAlgoSwitch option of the
AQM algorithm switch parameter selected to enable AQM. Then, enable the UE to re-access
the network.
Step 4 Download one or more large files and a small file simultaneously. Then, record the small file
download latency. Repeat the test and average the recorded values. The average value works
as the small file download latency with AQM enabled.
Step 5 Compare the two average values and verify that the small file download latency is shortened
after AQM is enabled.
----End
6.4.4 Reconfiguration
None
6.4.5 Deactivation
The following table provides the parameter used to deactivate this feature.
After AQM is enabled, the average processing delay of downlink PDCP SDUs for services
carried on DRBs possibly decreases slightly and the number of downlink PDCP SDUs
discarded for services carried on DRBs possibly increases slightly.
NOTE
The average processing delay of downlink PDCP SDUs for services carried on DRBs can be monitored
by dividing L.Traffic.DL.PktDelay.Time.QCI.x by L.Traffic.DL.PktDelay.Num.QCI.x, where x indicates
a QCI.
7 Parameters
CellAlg LocalCe DSP None None Meaning: Indicates the local ID of the cell. It uniquely
oSwitch llId CELLU identifies a cell within an eNodeB.
LCAMC GUI Value Range: 0~255
CLUST
ER Unit: None
8 Counters
9 Glossary
10 Reference Documents
None