Professional Documents
Culture Documents
Key Performance Guarantee Steps For UMTS Swapping V2.0
Key Performance Guarantee Steps For UMTS Swapping V2.0
Key Performance Guarantee Steps For UMTS Swapping V2.0
2013-09-30 www.huawei.com
Abstract
Abstract
Swapping Cases
Key Performance Guarantee Steps: risk evaluation and re-negotiation, intra-frequency solution design, KPI mapping, networking mapping, capacity
mapping, power mapping, parameter mapping, and key feature mapping
Abstract
As an example, if the DPA call drop rate is low, check difference between related features, for example, the DPA compressed
mode algorithm of the Nokia and Siemens Networks (NSN).
Evaluating the performance of the original network requires attention to the following points:
Difference that may cause KPIs to fail the requirement needs to be specially focused during key feature mapping and parameter mapping.
Record KPIs that have significant changes after the swapping during traffic analysis and ask the customer about the possible causes,
for example, whether a major network adjustment has been performed or a major event has taken place.
Seasonal changes or major events are considered to avoid KPI deterioration periods, for example, starting days of a new term for students.
If it is rather risky to meet the KPI requirements, seek help from the product support engineers.
Risk evaluation and re-negotiation are required throughout a swapping project. KPI risks need to be closely watched out during intra-frequency/inter-
frequency solution design, KPI mapping, parameter mapping, and other key steps.
KPI re-negotiation helps significantly reduce delivery risks in some cases. For example, a total of 20 plus acceptance KPIs were
narrowed down to only five in a project, which greatly eased delivery workloads and facilitated KPI acceptance.
Prepare site-specific intra-frequency/inter-frequency swapping solutions and Iur internetworking policies to lower KPI acceptance risks.
Before swapping, guide the customer to adopt Huawei-favorable KPI formulas specified in the KPI
mapping document. This helps facilitate KPI acceptance after swapping.
PS 0/0k
Packet Call
PS Call and Packet Call is different. Packet call failures can drop into FACH but no count failure in PS calls in the NSN
solution.
comparison.
Above failure statistical result does not include:
UE side failure: For example UE cannot send cell update message or RNC does not
receive cell update message, this case will cause abnormal release because of timer expiry
for example UE cannot select appropriate cell to send cell update message in weak
coverage area
Impact analysis:
RAB_ACT_FAIL_PS_XXX_XXX is used to count PS call drops, and abnormal PS releases in some scenarios are not
counted, which helps reduce the PS call drop rate. The Packet Call parameter of Huawei can more accurately reflect
the PS call drop rate of the DCH. The following is an example of calculating the PS call drop rate by using the packet
call:
After the NSN related failure counter is considered, the PS call drop rate reaches 0.99% instead of 0.11%.
The RRC connection release measurement (RRC.Rel.Cell) differs from the NSN RRC call drop measurement in statistical dimensions in the following
aspects:
1. The NSN network measures only CS abnormal releases, but the Huawei network measures both CS and PS abnormal releases.
2. Some CS abnormal releases are not measured on the NSN network. As shown in the figure, CS abnormal releases caused by SRB resets triggered by
an RB setup procedure (red circled) and by failed safe mode procedures are no counted. Huawei networks measure all CS and PS abnormal releases
as a whole through the RRC.Rel.Cell counter, without separately counting CS abnormal releases.
Impact analysis and workaround:
The mapping of this KPI between Huawei and NSN cannot be performed. The measurement of this KPI of NSN does not make sense and is useless in
practice. As shown in the red circle, the RRC call drop caused by an SRB reset due to an RB SETUP message delivered from the network side after a
successful RRC connection establishment is not counted. In this case, explain to the customer that the mapping failure is caused by insufficient
measurement of the RRC call drops of the NSN and encourage the customer not to use the RRC call drop rate for swapping acceptance and network
monitoring.
PS Call and Packet Call is different according to NSN, Packet Call failure can drop into FACH but no count failure in
PS Call
Packet Call
The tool supports automatic recognition and error correction of the user camping policy and
service bearer policy of a peer-vendor network. For example, if mutually preferential camping
policy between carriers is adopted on a network built by other vendors, ping-pong reselection
will occur, and power consumption and access success rate will be greatly affected after the
network is swapped. This tool can help swapping engineers recognize and notify them of
such problems.
The SmartRNO supports use of peer-vendor scripts to identify the peer-vendor networking policy.
1. Collect peer-vendor
Covers
counter data. \tab 2. Sort out the 3. Import the data.
transmitted Mapping of different data.
carrier power resources requires different Sort out and analyze the data
collected in Step 1 through
(TCP), counter data and methods
the following templates. After
channel for obtaining counters. It is
the collected data is added to
element (CE), good practice to collect data the templates, the tool
Common of the original network
directly uses data in the
NodeB generated within the latest
template for mapping.
Application seven days.
Protocol
(CNBAP), 5. Check the result data.
signaling 4. Perform capacity mapping.
processing
unit (SPU),
data
processing
unit (DPU),
and interface
boards (INT)
This solution uses a capacity mapping tool to calculate the capacity requirements and provide capacity mapping solutions based on traffic volume
and models of the peer-vendor networks. This helps ensure optimum resource configuration and minimum swapping risks for the swapping.
Power mapping mainly deals with the mapping of cell power, pilot and common channel power, and radio link power
between the peer vendor and Huawei.
The Parameter Mapping Table covers load control parameters, power control parameters, channel configuration
parameters, mobility management parameters, terminal compatibility parameters, NE internetworking parameters, new
feature parameters, and methods for matching these parameters. The Parameter Mapping Table uses the following
format:
Mapping description:
Policy Policy description
Recommended: Parameters of this type need to be mapped before swapping by using the mapping rules.
Mandatory Parameters of this type need to be mapped before swapping by using the recommended optimization value.
Optional Parameters of this type need to be mapped before swapping based on the risk evaluation results.
Priority description:
Principle:
1. Parameters for cell selection or reselection of UEs in idle mode and parameters for timers, power control, soft handovers, inter-frequency hard handovers, inter-
RAT handovers, and state transition of UEs in idle and connected modes require high mapping priority and need to be specially focused.
2. Others that need to be explained to the customer before swapping:
The NSN inter-RAT handover mechanism is more likely to trigger inter-RAT handovers than Huawei's (for details, see the description about NSN inter-RAT
handover mechanisms). The number of inter-RAT handovers may decrease after swapping. The workaround method is to increase the threshold of events 2D
and 2F by one to two dB.
The Ericsson inter-RAT handover mechanism (based on event measurement) is less likely to trigger inter-RAT handovers than the inter-RAT handover
mechanism of Huawei (based on periodic measurement). This will cause the UMTS network to carry less traffic and the number of inter-RAT handover may
increase as well. The workaround method is to decrease the threshold of events 2D and 2F by one to two dB.
As an example, after the RU10 of NSN was swapped for operator WIND in Greece and Italy, a large number of neighboring cells were not
configured, which fails the call drop rate requirements. After other factors that could lead to the problem were excluded, the neighboring
cells were then added and the call drop rate increased to a level comparable to the original network.
Impact analysis:
Positive gains Negative gains
Automatic neighboring cell addition Improve call drop rates for CS services and other KPIs. Not found
Workaround:
1. It is good practice to use the SmartRNO to map the neighboring cells of the original NSN network.
2. Use a neighboring cell tool, for example, U-Net, to supplement the neighboring cell configurations before the swapping starts.
3. Use the Nastar to optimize the neighboring cell configurations after the swapping is complete.
4. Provide automatic neighboring cell addition in RAN15.0 of Huawei.
The Huawei network supports compressed mode for HSDPA services by default, and therefore HSDPA UEs do not trigger
event 1F and fall back to the DCH. As a result, the HSDPA KPIs will easily represent huge difference before and after
swapping.
If the NSN network is configured not to support compressed mode for HSUPA services, UEs in the weak coverage areas are
likely to trigger event 1F, and the HSDPA UEs consequently fall back to the DCH. This decreases the handover success rate
for HSUPA services and other network KPIs, as shown in the following table.
HSUPA HSUPA HSUPA HSUPA Call Drop User
KPI
Throughput Traffic Users Rate experience
Gain
√ × × √ ×
Evaluation
The Huawei network supports compressed mode for HSUPA services by default, and therefore HSUPA UEs do not trigger
event 1F and fall back to the DCH. As a result, the HSDPA KPIs will easily represent huge difference before and after
swapping.
√ denotes a positive gain and ×denotes a negative gain.
Workaround:
The NSN solution uses T314 and T315 to control call reestablishment. On the original NSN network, if T314 is not set to 0
and RU10 or later is used, active call reestablishment for CS services may be enabled. If T315 is not set to 0 and RAS06 or
later is used, active call reestablishment for PS services may be enabled. If this function is enabled on the original NSN
network, use the following two workarounds:
1. Persuade the customer to disable this function before measuring the benchmark KPIs of the original network. This is
because active call reestablishment not only prolongs mute duration but also covers a real network problem, which affects
network optimizations.
2. HUAWEI support active reestablishment and UE SRB reset reestablishment function, please follow Features documents
to active this function. During network swapping, the Parameter Mapping Table (Ericsson W11, NSNRU20, and ALU
UA5.1-Huawei V900R014_20121031) is required to optimize the configurations of the related parameters.
Points to consider when enabling the active call reestablishment are as follows:
In RN5.0 of NSN RU20, the active call reestablishment time is controlled by T313, T314, and T313. If the customer insists on
the use of active call reestablishment in a swapping project, the active reestablishment timers of Huawei and NSN generally
need to be mapped. The NSN timer (generally about 8s) is shorter than Huawei timer. Simulation tests show that NSN uses
active call reestablishment in more scenarios than Huawei. Therefore, the NSN timer parameters are mapped directly and
the call drop rate may increase after swapping. In this sense, direct mapping is not recommended. For details about the
timer settings, see the Call Re-establishment sheet in the parameter mapping table (Ericsson W11, NSNRU20, and ALU
UA5.1-Huawei V900R014_20121031).
Abstract
Swapping Cases
U-SwappingPerf
ormanceGuaranteeTemplate
Abstract
Swapping Cases
Case 1: PS RAB setup success Rate Failed to Meet the Requirement Caused by Incomplete Mapping
Due to KPI Algorithm Difference
Problem description:
In swapping project P, parameter mapping is required between Huawei and NSN. The KPI data provided by the customer
shows that the PS RAB setup success rate of the original NSN network stands as high as 99.9x%. This puts acceptance of
the swapped network at great risks.
Cause analysis:
The original network uses the NSN RU 20 in which the PS 0k/0k feature is already incorporated. The PS RAB setup
success rate provided by the customer is actually obtained after the PS 0k/0k feature is used.
sum([RAB_ACC_COMP_PS_INTER] + [RAB_ACC_COMP_PS_BACKG])/sum([RAB_STP_ATT_PS_INTER] +
[RAB_STP_ATT_PS_BACKG])
In opposition to the PS RAB setup success rate measured by Huawei, this KPI measured by the NSN does not reflect the
true PS RAB setup success rate. Therefore, a modification of the above formula to the following one is recommended:
<RAB setup success rate of PS services with PS 0k/0k enabled>* <RB reconfiguration success rate>
Impact analysis:
The PS RAB setup success rate used by NSN and Huawei has different meanings. If mapping were performed by using
the NSN formula, this KPI would fail to meet the requirement after the network is swapped.
Solution:
Explain to the customer the real meaning of the PS RAB setup success rate measured by NSN and Persuade the
customer to allow mapping based on the modified formula. The customer agreed the mapping based on the modified
formula and to allow a certain margin during acceptance of this KPI.
Case 2: PS RAB setup success Rate Failed to Meet the Requirement Caused by Incomplete Mapping
Impact analysis:
The RRC-based DRD policy adopted by the NSN is favorable to the PS RAB setup success rate that cannot
be achieved by the RAB DRD policy used by Huawei.
Solution:
Protective measures for possible RAB DRDs are adopted for the swapped network. If a DRD failure occurs
during the RB setup procedure, DRD rollback is performed and the RB setup is performed again, and the
first DRD failure is not measured as an access failure. This helps increase the PS RAB setup success rate.
Check the status of the DRD rollback switch on the swapped network, finding that this switch is not turned on
(which is inconsistent with the swapping SOP).
The DRD rollback during the RB setup procedure is then enabled. The PS RAB establishment rate increases noticeably
and is even higher than that on the original network.
Case 5: RRC Establishment Success Rate Failed to Meet the Requirement Due to Algorithm Difference
in Handling Abnormal Terminals
Problem description:
During an NSN network swapping in country N, the onsite engineer reports wide fluctuation of the RRC
access success rate, with low values in certain periods of time, putting the acceptance at risks.
Cause analysis:
The PCHR analysis for the network shows that there are TOP N terminals that keep sending the RRC
CONNECT REQ messages without responding to the RRC SETUP messages from the network side. As a
result, the RRC success rate stays at a low level at the site with such terminals.
Impact analysis:
The algorithm for handling such abnormal terminals helps increase the RRC establishment success rate.
Solution:
Explain to the customer the difference in algorithm for handling abnormal terminals and promise a solution timetable for
this scenario to the customer.
Cause analysis:
The calculation formula is not appropriate
because the call drop rate in some scenarios is
not counted.
Impact analysis:
This affects the acceptance of PS call drop
rate of the swapped network.
Solution:
Persuade the customer to agree to calculate
the call drop rate through the Huawei-provided
formula.
Impact analysis:
The call drop rate of HSUPA services increases in the weak coverage area of the swapped network.
Solution:
Explain to the customer the algorithm difference and that the PS call drop rate actually decreases on the swapped
network.
HUAWEI TECHNOLOGIES CO., LTD. Huawei Confidential Page 42
Swapping Case: Call Drop in Parameter Mapping
Case 8: Call Drop Rate Increased Due to Algorithm Difference in Composite Service Bearer Policy
Problem description:
After an NSN network is swapped, the CS call drop rate increases in multiple clusters as compared to the original network.
Cause analysis:
On the NSN network, the rate of composite services is reset to 0 K byte in peak hours and to 384 K bytes in off-peak
hours. It is assumed that certain function is enabled to limit the rate of PS services based on cell loads. After the swapping
is finished, channel fallback for composite services and rate limiting are not used.
Impact analysis:
The difference in bearer policy of composite services leads to increased CS call drop rate after swapping.
Solution:
The bearer policy of composite services is changed onsite, and the CS call drop rate decreases by 0.09%.
Impact analysis:
Due to improper power mapping, the transmit power of cells increases, which pushes up the PS traffic. As a result, the
CS and PS call drop rates are affected.
Solution:
Match the pilot power and cell power of the original network.
Case 11: HSDPA MAC-hs Efficiency Decreased After the Swapping Caused by Incomplete Parameter
Mapping Due to KPI System Difference
Problem description:
After an Ericsson network is swapped in a project, the HSDPA MAC-hs efficiency is about 10% lower than the original
Ericsson network.
Cause analysis:
The statistic scenario of this counter is different between Huawei and Ericsson, which is unfavorable to the counter values
of Huawei.
Formula of Ericsson:
(100% x pmNoActiveSubFramesSpiXX)/(pmNoActiveSubFramesSpiXX + pmNoInactiveRequiredSubFramesSpiXX)
In this formula, an XX stands for an SPI ranging from 00 to 15.
Formula of Huawei:
(VS.DataTtiRatio.Mean –VS.HSDPA.InactiveDataTtiRatio.Mean)/VS.DataTtiRatio.Mean
In the Ericsson formula, the pmNoInactiveRequiredSubFrames counter only measures HSDPA UE queues in the
scheduling candidate set. The MAC-hs scheduler does not schedule the HSDPA UEs in the 16 scenarios (identified
together with the customer). Therefore, these UEs are not measured for calculating the MAC-hs HSUPA efficiency, among
which the HSUPA UEs with unreliable uplink quality are included.
Case 11: HSDPA MAC-hs Efficiency Decreased After the Swapping Caused by Incomplete Parameter
Mapping Due to KPI System Difference
Impact analysis:
In RAN13.0, the ratio of the time when at least one HSDPA UE in a cell has data stored to transmit in the queue buffer to
the total time within a measurement period is measured, and the HSUPA UEs that are not in the scheduling candidate set
are counted. Therefore, the statistical method is unfavorable to Huawei. Analysis of the live network data shows that above
80% HSUPA UEs are not scheduled due to unreliable uplink quality.
Solutions:
1. Explain to the customer the difference and perform KPI mapping by strictly following the swapping SOP.
2. The VS.HSDPA.ScheInactiveDataTtiRatio.Mean counter is added and the following formula is used in RAN14.0 to match
the Ericsson calculation method.
HSDPA efficiency = (VS.DataTtiRatio.Mean – VS.HSDPA.InactiveDataTtiRatio.Mean)/(VS.DataTtiRatio.Mean –
VS.HSDPA.InactiveDataTtiRatio.Mean + VS.HSDPA.ScheInactiveDataTtiRatio.Mean)
Case 12: Retransmission Rate on the Air Interface for HSDPA Services Increased After Swapping Due to
Algorithm Difference
Impact analysis:
The TFRC policy adopted by Ericsson is more favorable to control the BLER, but may be not the most favorable to the
throughput.
Solutions:
1. Explain to the customer that the difference in HSDPA retransmission rate is caused by the TFRC algorithm difference.
2. Explain to the customer that the BLER is an intermediate counter and that a high HSDPA retransmission rate does not
necessarily mean low HSDPA throughput. In this sense, a direct HSDPA retransmission comparison has little bearing. It
is good practice to ask the customer to focus on the counter that directly boils downs to the ultimate result.
Case 13: HSDPA Throughput Failed to Meet the Requirement Due to Incomplete Parameter Mapping and
Algorithm Difference
Problem description:
After an ALU network is swapped in project V, the HSDPA throughput of cells in above 50% clusters is decreased. In some
clusters, the difference goes as far as about 10%.
Cause analysis:
1. Parameter mapping
The HSDPA parameters for the dual-carrier networking are not appropriately mapped (the HSDPA is disabled in the
cell using carrier F1 on the original network, but is enabled on the swapped network.
The power mapping does not strictly follow the cell-level mapping principle (the Alcatel-Lucent uses the antenna port
as the power reference point, yet Huawei uses the cabinet top as the power reference point).
2. Algorithm difference:
Difference in the State Transition Algorithm
The customer-provided information demonstrates that the Alcatel-Lucent uses the throughput as a criterion
for state transition of UEs. Following is an excerpt of the customer-provided information.
When UE is in the DCH state, RNC monitors UL and DL throughput with sliding window of 500 milliseconds; if
throughput is below 8 kbit/s, low activity is pointed out; if low activity persists for 10 sec in case of DCH-DCH, 16 sec in
case of DCH-HSDPA and 25 sec in case of EDCH-HSDPA, RNC reconfigures UE to cell FACH state.
Case 13: HSDPA Throughput Failed to Meet the Requirement Due to Incomplete Parameter Mapping and
Algorithm Difference
Cause Analysis (continued):
Algorithm difference:
HSDPA remaining power appending algorithm (not available in RAN14.0)
RLC processing difference at Layer 2
With the RLC processing mechanism used by Huawei, if RLC polling is triggered before the POLL Prohibit Timer expires,
a packet data unit (PDU) is immediately retransmitted and the bit is set to POLL after the POLL Prohibit Timer expires if
the RLC polling is not confirmed. The RLC processing mechanism of Alcatel-Lucent rules that a POLL bit is directly
attached to a PDU, regardless of the status of the POLL Prohibit Timer.
Solutions:
1. Parameter mapping is performed by strictly following the principle of cell-level mapping before the swapping.
2. Explain to the customer the difference in the state transition and HSDPA remaining power appending algorithms.
Optimize the parameter configurations of the swapped network to match the original network as much as possible (1%-
3%) and promise to provide similar algorithms in a future version.
3. Explain to the customer the difference in RLC processing mechanism. Disable the POLL Prohibit Timer to match the
original network configurations as much as possible (2%-4%).
The preceding methods can be used to help the cell HSDPA throughput to meet the requirement.
Case 14: Number of HSUPA UEs Increased Sharply After Swapping Due to Incomplete Parameter
Mapping and Algorithm Difference
Problem description:
After an Ericsson network is swapped in project N, the number of HSUPA UEs increases by about 70% based on cluster-
level statistics.
Cause analysis:
1. The maximum number of HSUPA UEs is set to 16 on the original network (maxNumEulUsers), but it is set to 40 on the
swapped network (MaxHsupaUserNum).
2. The trigger mechanism for state transition from HS-DSCH to FACH is difference before and after swapping. For details,
see the Algorithm Difference section.
3. The state transition timer parameters (BeH2FStateTransTimer and BeE2FStateTransTimer) are mapped to the original
network.
Impact analysis:
A sharp increase in the number of HSUPA UEs results in excessive uplink resources, which imposes excessive heavy
loads on the uplink and consequently affects the network KPIs and HSUPA throughput.
Solutions:
1. Perform parameter mapping according to the swapping SOP to match the parameter configurations of the original
network as much as possible.
This helps make the number of HSUPA UEs on the swapped network comparable to the original network.
Solution:
The KPI formula used by NSN is modified so that the inter-RAT handover KPIs are successfully accepted.