LTE Downlink Throughput Optimization Based On Performance Data
LTE Downlink Throughput Optimization Based On Performance Data
www.huawei.com
2 Technical Principles
4 Tools
5 Delivery Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 2
LTE Large-Scale Commercial Application, Fast Network Construction
The commercial application of LTE networks speeds up; the terminal industrial chain matures in
2014; and the number of UEs increased dramatically.
The number of commercial LTE networks has reached 274 (28 LTE
UE Quantity 5 million 10 million 100 million
TDD networks) by February 17, 2014
300 263 274
2 years 3 years 6 years
250 2G
(1992–1994) (1992–1995) (1992–1998)
200
148 2 years 3 years 6 years
150 3G
(2001–2003) (2001–2004) (2001–2006)
100
46 1.8 years 2 years 3.3 years
50 2 16 4G
(Jan 2010–Oct 2011) (Jan 2010–Jan 2012) (Jan 2010–Apr 2013)
0
2009 2010 2011 2012 2013 2014.2 The number of LTE users exceed 100 million in 3.3 years, doubling
the growth rate of 2G or 3G UEs.
According to GSA, the number of TD-LTE UE types has increased to
304 by January 30, 2014, among which 63 were smart UEs,
exceeding 20% of the total number. The LTE TDD UE industry chain
is mature. Increasing rapidly
since 2012
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 3
Insufficient LTE Network Optimization Capability
Issue 2: Optimization bottlenecks caused by limited optimization
Issue 1: As an end-to-end network KPI, LTE throughput may
methods due to insufficient in-depth analysis capability
fluctuate due to various factors.
of RF problems in LTE networks
Capacity
expansion
Neighboring cell
Building new
sites
Bottleneck!
Issue 3: Poor batch delivery and high labor cost due to the current Issue 4: High network risk and O&M cost due to experience-based
optimization platform which provides only process data for adjustment or uniform optimization using baseline values
analysis, but no clear analysis results and closure actions. Suggestions
not
differentiated
SmartRNO:
1. Supporting neighboring cell planning Increase the Value on
Cell Baseline Optimization
value as an the Live
checks (topology analysis only) Name
Network
Value Suggestion
attempt. If
FMA: KPIs Cell A X1 Y1 X1–Y1
1. Supporting NE fault analysis and top deteriorate,
cell filtering decrease the Cell B X1 Y1 X2–Y1
FMA 2. Cannot provide optimization value again.
Cell C X2 Y1 X2–Y1
suggestions
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 4
Contents
1 Background
2 Technical Principles
4 Tools
5 Delivery Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 5
Optimization Scheme
Low spectral
Low throughput Poor user experience
efficiency
Identification of
KPIs evaluation and Identification problematic cells Select top cell groups based on
of problematic cells
customer concerns and focus on
MRF top sites
Key parameter adjustment PCI adjustment Channel check ACP antenna adjustment
Optimization suggestion for
different root causes
Optimization
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 6
Top Cell Identification
Select top N (configurable) cells that have the worst performance indicated by the specified KPIs
Cell KPIs such as throughput, channel quality, MCS level, rank 2 ratio, and BLER are displayed.
Use the total duration of downlink data transmission in a cell to calculate a more accurate average cell throughput in the downlink.
Calculate the downlink bearer throughput per UE according to 3GPP specifications and exclude the traffic in the last TTI to obtain a more
accurate result.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 7
End-to-End Problem Isolation by Segment
Air interface Root Cause Non-Air Interface Public Network Root Cause
Root Cause
Air interface channels: Factors 1. NEs with flow
1. Incorrect parameter 1. Incorrect TCP
1. Air interface encoding Bearer network 1. Incorrect control
settings parameter settings
(MCS/MIMO/IBLER) transmission channels: parameter settings 2. Public network
2. Limited traffic capacity 2. Capacity
2. Air interface resources 1. Bandwidth restriction 2. Limited capacity bandwidth
3. Poor coverage restrictions
(Grant/RB) 2. Long delay and jitter or capability restrictions
4. Interference
3. Weak 3. Packet loss and 3. Transmission
5. Handover exceptions
coverage/interference disorder quality problems
1
Uu
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 8
Root Cause Tree of LTE Throughput Problems
Problem Type Root Cause Analysis and Troubleshooting Method Automatic Analysis Applicable Remarks
If possible, connect the test PC directly to the SP server to test low throughput cells
Information about SP servers in commercial use cannot be obtained. Such tests can be
Server capability where a certain service is used by a large number of UEs or the download speed of a No
performed only on FTP servers or servers for China Mobile-owned services.
certain website is slow.
SP server
TCP parameters Check the TCP window size. No None
Software configurations Enable the firewall. SP servers do not support TCP port extension. No None
NEs with flow control Check DPI nodes. The GGSN communicates with the OCS at a too high frequency. No None
Public network
Public bandwidth limitation None No None
Registration configurations Analyze MME CHR data. No, but technically feasible. MMEs must have MV capability for analysis if they are provided by other vendors.
Rate limitation Analyze MME CHR data. No, but technically feasible. None
Packet disorder Analyze MME CHR data. No, but technically feasible. None
Check whether such counters as static resource usage, signaling processing resource
Insufficient equipment No. The capacity depends on usage, forwarding resource usage (SAU/PDP), forwarding bandwidth usage (Mbit/s or
Check the CN equipment capacity usage.
Core network capacity traffic statistics on the core network. pps), comprehensive resource usage, and average CPU usage are available on the CN
side.
Incorrect parameter No. The capacity depends on MML
Set core network parameters. None
settings commands on the core network.
Software defects This problem only occurs in the area covered by equipment provided by a sole vender. No None
Equipment incompatibility This problem only occurs in the area covered by equipment provided by a sole vender. No None
Prerequisites of performing such analyses:
Bandwidth limitation Analyze IPPM and GTPU counters. Yes 1. IPPM counters must be available.
2. The core network equipment in use must be provided by Huawei.
Prerequisites of performing such analyses:
Long delay and large jitter Analyze IPPM and GTPU counters. Yes 1. IPPM counters must be available.
2. The core network equipment in use must be provided by Huawei.
Prerequisites of performing such analyses:
Packet loss and disorder Analyze IPPM and GTPU counters. Yes 1. IPPM counters must be available.
Transmission 2. The core network equipment in use must be provided by Huawei.
Check whether there are multiple eNodeBs in the same access ring by querying the XML
Too many eNodeBs in the Check the bandwidth usage of the eNodeB access ring. (This counter is unavailable
No files of the eNodeB. You can perform the check manually, because automatic analyses
same access ring on the wireless network side.)
may be inaccurate due to inaccurate data sources.
Bandwidths occupied by Check the bandwidth usage of the eNodeB access ring. (This counter is unavailable
No This is a counter on the transmission side, which may be unavailable.
services on the access ring on the wireless network side.)
Incorrect UGW Check the quintet configuration and related feature switches. Packets are lost due to
No Information about UGW configurations is needed for such analysis.
configurations high instant rate of the UGW output port.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 9
Root Cause Tree of LTE Throughput Problems
Automatic
Problem
Root Cause Analysis and Troubleshooting Method Analysis Remarks
Type
Applicable
eNodeBs fail or their capability The alarm list was complete and consistent with that of X Solutions for PS
Analyze relevant alarms. Yes
degrades Data Transmission Optimization in the LTE Network.
eNodeB rate restriction Check the licensed traffic volume. Yes This method is not verified.
Analyze traffic statistics (Check whether the CPU usage is The analysis results of the CPU usage counter were consistent with that of X
eNodeB processing capability Yes
NE excessive). Solutions for PS Data Transmission Optimization in the LTE Network.
Check known problems. Check whether features have any gain. Part of the analysis results were verified. (Analysis results of MLB were
Algorithm and feature restrictions Yes
Check the strategies for enabling features. verified in Hangzhou.)
The analysis results of these usage counters were consistent with that of X
eNodeB capacity restricted Check the usages of CCE, SRS, PUCCH, and PRB resources. Yes
Solutions for PS Data Transmission Optimization in the LTE Network.
Antenna faults and alarms Analyze relevant alarms. Yes None
Invoke the
Reverse connection or connection
Another document is dedicated to this problem. corresponding None
with crossed pairs for antennas
Antenna document.
Uplink interference Another document is dedicated to this problem. Yes The analysis results are accurate, which have been verified in Hangzhou.
Check whether there is path loss imbalance of dual-channel
Imbalanced channels No None
components in indoor distributed eNodeBs with two antennas.
Air interface resources (Grant/RB) Analyze traffic statistics. Yes None
Air
Weak coverage Analyze the measurement report. Yes The MR analysis results are accurate, which have been verified in Hangzhou.
interface
Interference Analyze the measurement report. Yes The MR analysis results are accurate, which have been verified in Hangzhou.
UE capability Analyze CHR signaling. Yes The analysis results are accurate, which have been verified in Hangzhou.
QCI configurations Analyze CHR signaling. Yes The analysis results are accurate, which have been verified in Hangzhou.
UE AMBR rate Analyze CHR signaling. Yes The analysis results are accurate, which have been verified in Hangzhou.
UE main and diversity reception Perform UE trace test, because a background analysis is not Data sources for such analysis are unavailable and the analysis results
No
imbalance available. cannot be verified.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 10
End-to-End Problem Isolation by Segment
and capacity
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 11
Weak Coverage Root Cause Location
Weak coverage scenarios:
Missing configurations of neighboring cells/delayed
handovers/island coverage/no primary servicing cells/link
imbalance/weak intensive coverage
Serving cell Target cell
Delayed
handovers UE
Missing Low RSRP of the serving cell and high RSRP of the target cell.
No primary
configurations
of neighboring
serving Unconfigured neighbor relationships: If the distance between the UE
cells
cells
and the serving eNodeB is smaller than the average inter-site
Identification of distance, the root cause of low throughput is missing configuration of
weak coverage neighboring cells.
root causes
If the distance between the UE and the serving cell is greater than the
average inter-site distance, the root cause of low throughput is island
Island Link
coverage imbalance coverage.
If neighbor relationships are configured and the root cause of low
throughput is delayed handover, adjust handover parameters based
Weak
intensive on the signal level difference between the neighboring cell and the
coverage
serving cell.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 12
Weak Coverage Root Cause Location
-104 dBm
-102 dBm
Serving cell UE Serving cell
Neighboring cell 1
UEs close to the serving eNodeB are in weak coverage areas. This
phenomenon is called coverage hole. This occurs because of a coverage
hole, mostly in indoor scenarios close to eNodeBs UE
-101 dBm
Neighboring cell 2
Serving cell
UE
At the coverage border of multiple neighboring cells that have small
Due to link reciprocity in LTE TDD networks, path loss imbalance occurs difference in the signal level, there is no primary serving cells with strong
between the uplink and downlink. As a result, downlink signals may be too signals, increasing the probability of ping-pong handovers.
weak for UEs far from eNodeBs to receive, causing link exceptions. This root
cause does not lead to problems to LTE FDD cells.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 13
Interference Root Cause Location
Interference scenarios:
overshoot coverage/PCI mod 3 conflicts/pilot
pollution/azimuth exceptions
Serving cell
Neighboring cell 1
Pilot
pollution
UE
The RSRP of the serving cell is high, but the channel quality is low.
If the serving cell has an intra-frequency neighboring cell with a similar
RSRP value, intra-frequency interference has a large impact on the
throughput. If PCI mod 3 conflicts occur, the SINR deteriorates.
If the serving cell has no intra-frequency neighboring cells or its
PCI mod 3 neighboring cells have low RSRP, inter-RAT interference may occur,
conflicts
such as intermodulation interference of GSM 1800 band to band F and
outband interference from PAS.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 14
Interference Root Cause Location
Method of checking the azimuths of cells (relative locations between cells)
PCI: 0
PCI: 6
-80 dBm -84 dBm
If PCI mod 3 conflicts occur between the serving cell and neighboring cells and
the reception level of the serving cell is lower than 6 dB, interference occurs in
pilot channels. As a result, the SINR on the RS channel deteriorates, and intra-
frequency PCI mod 3 conflicts occur, affecting the throughput. Interference due to azimuth exceptions occurs if the following conditions are met at the
same time:
• Relative locations between the serving cell and neighboring cells do not cause
overlapping coverage.
• UEs receive signals from the neighboring cells.
• PCI mod 3 conflicts occur.
• The reception level of the serving cell is lower than 6 dB.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 15
Interference Root Cause Location
Neighboring cell PCI: 6
PCI: 0 PCI: 0
-84 dBm PCI: 1
-80 dBm
-80 dBm -83 dBm
-90 dBm
-91 dBm
If PCI mod 3 conflicts do not occur in multiple cells but
Interference the RSRP of these cells are close, there is a probability
Serving cell that carriers for data service transmission collide with
cell
UE the RS channel, causing interference to the carriers. As
Overshoot coverage of the serving cell occurs when signals from a result, the SINR deteriorates, and the throughput of
the serving cell passing through neighboring cells in between the serving cell decreases, causing pilot pollution.
cause overlapping coverage with a remote cell.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 16
Accurate Location of Un Interface Problems
Calculate the number of RF root causes by cell and obtain the distribution of weak coverage and interference root cause percentages
Cause classification of weak coverage problems
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 17
Root Cause-based Optimization Suggestions
Parameter classification for optimization in specific problematic scenarios
1. Find the exact missing
1. Identify incorrect handover
1. Provide optimization 1. Provide accurate frequency Provide targeted optimization
parameter settings based on air 1. Check and optimize channels
neighbor relationships, suggestions on antenna optimization suggestions
interface quality and TMAs to address the
instead of performing
2. Perform targeted optimization for problem of severe path loss
downtilts for cells based on interference methods for different problems to
topology-based search experiencing overshoot measurement results and
different handover parameters imbalance between the uplink
2. Quantize handover
parameters (CIO)
(magnetic hysteresis parameters and downlink
coverage or weak intensive
coverage
changes associated with PCI
optimization
ensure optimization effect.
and CIO)
Perform intensive parameter
management to lower the
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 18
Analysis Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 19
Contents
1 Background
2 Technical Principles
4 Tools
5 Delivery Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 20
Introduction to Themes on the OMStar
Tool: OMStar V500R012C00
Download address:http://support.huawei.com/carrier/navi#col=software&allProduct=true&path=PBI1-7275736/PBI1-9855723/PBI1-7497478/PBI1-59728
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 21
Introduction to Themes on the OMStar
4. Import data.
Data Source (Data Marked
with Asterisks Are Specifications
Mandatory.)
* External CHR data (SIG) 8 GB
*Performance data (MRF) 50 MB
High
*Configuration file (XML) 500 MB
efficiency in
importing NIC *Engineering parameters
20 MB
packages (XLS/XLSX)
Operation log 500 MB
Alarm log (CSV) 50 MB
License file (TXT) 50 MB
Number of cells 300
6 busy hours in
three days
Data collection period
(3 days x 2 busy
hours)
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 22
Introduction to Themes on the OMStar
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 23
Introduction to Themes on the OMStar
7. Set required parameters. 8. Select the time period of performance data for analysis.
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 24
Introduction to Themes on the OMStar
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 25
Contents
1 Background
2 Technical Principles
3 Tools
4 Delivery Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 26
LTE Throughput Rate Optimization Delivery Guide
• Input
Theme-related data sources (XML configuration data, engineering parameters, MRF
performance data, SIG file, alarm data, operation log, and license query record)
Parameter configuration rules
Description of engineering parameter fields
Automatic analysis on the OMStar
Field Name Description Remarks
• Output eNodeBID Indicates the ID of an eNodeB. Mandatory
TopN cell evaluation reports Cellid Indicates the ID of a cell. Mandatory
Problem analysis reports PCI Indicates the physical cell ID. Mandatory
CellName Indicate the name of a cell. Optional
Optimization suggestions Uarfcn Indicates the cell frequency. Mandatory
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 27
Optimization Report Interpretation
UE UE Problem Advice
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 28
Optimization Report Interpretation
Mapping between root cause locating results and optimization suggestions
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 29
Contents
1 Background
2 Technical Principles
3 Tools
4 Delivery Process
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 30
Appendix 1: Throughput Rate Calculation Principles
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 31
Appendix 1: Throughput Rate Calculation Principles
• Broadcast channel
› PBCH: The first four OFDM symbols for each wireless frame with slot No. 1 are obtained: [(4 x 72)/(14 x 12 x 100)]/10 ms =
0.2%
› The SIB-1 period is 20 ms, the number of occupied RBs is 13, and the occupation overhead is calculated: (13/100)/20 ms =
0.65%
› The SI-1 period is 40 ms, the number of occupied RBs is 12, and the occupation overhead is calculated: (12/100)/40 ms =
0.3%
› The SI-2 period is 80 ms, the number of occupied RBs is 19, and the occupation overhead is calculated: (19/100)/80 ms =
0.24%
› The SI-3 period is 80 ms, the number of occupied RBs is 41, and the occupation overhead is calculated: (41/100)/80 ms =
0.51%
› The SI-4 period is 160 ms, the number of occupied RBs is 32, and the occupation overhead is calculated: (32/100)/160 ms
= 0.2%
› The SI-5 period is 160 ms, the number of occupied RBs is 32, and the occupation overhead is calculated: (32/100)/160 ms
= 0.2%
› Delivering information about SI from 1 to 5 is optional. The half overhead value is calculated and SIB-1 is added so the total
overhead is 1.375%.
• FDD peak value
› If the coding rate is 1 and 64QAM modulation mode used (each RE occupies 6 bits), the peak rate is calculated using the
following formula in 20 MHz LTE signal 2 x 2 MIMO: [100 x 12 x 14 x (1 – 9.5% – 0.2% – 12% – 0.17% – 0.2% – 1.375%) x
6 x 2]/1 ms = 154.33 Mbit/s
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 32
Attached RS R0 R0
R0 R0
R0 R0
l0 l6 l0 l6
R0 R0 R1 R1
R0 R0 R1 R1
Reference symbols on this antenna port
R0 R0 R1 R1
l0 l6 l0 l6 l0 l6 l0 l6
R0 R0 R1 R1 R2 R3
Four antenna ports
R0 R0 R1 R1 R2 R3
R0 R0 R1 R1 R2 R3
R0 R0 R1 R1 R2 R3
l0 l6 l0 l6 l0 l6 l0 l6 l0 l6 l0 l6 l0 l6 l0 l6
even-numbered slots odd-numbered slots even-numbered slots odd-numbered slots even-numbered slots odd-numbered slots even-numbered slots odd-numbered slots
HUAWEISEMICONDUCTOR
HISILICON TECHNOLOGIES CO., LTD. HUAWEI Confidential Page 33
Thank you
www.huawei.com