Escolar Documentos
Profissional Documentos
Cultura Documentos
In B9, transmission resource management algorithms were enhanced with respect to B8 in order to optimise resource usage at Abis and Ater interface level.
M-EGCH Statistical Multiplexing Dynamic Abis Allocation Ater resource management DL retransmission in BTS
(1/5)
This feature provides a solution to share the Ater and Abis nibbles between the radio timeslots of a TRX so that the transmission resources left available by a PDCH can be reused by other PDCHs as long as those PDCHs belong to the same TRX. An M-EGCH link (Multiplexed Enhanced GPRS CHannel) is a bi-directional link established between the MFS and the BTS.
CS traffic
7 6 5 4
TCH TCH
GCH Basic
3
PS traffic
2 1 0 TRX
M-EGCH link
GCH Extra
(2/5)
1
EGCH
2
EGCH
3
EGCH
4
EGCH
5
EGCH
6
EGCH
TRX
M-EGCH link
composed of
GCH
composed of
GCH GCH GCH
GCH
GCH
GCH
GCH
1 to 36 GCHs
GCH
GCH
(3/5)
(4/5)
(5/5)
The size of an M-EGCH link (associated to a TRX) can be dynamically decreased or increased. The algorithms to dynamically decrease or increase the size of an M-EGCH link correspond to the Dynamic Abis allocation.
(1/5)
This feature enables to dynamically allocate Abis nibbles among the different TRXs used for PS traffic in a given BTS. Compared to B8, it allows a higher average Abis bandwidth per PDCH, the BSC capacity in terms of TRXs is increased, and in some BTS configurations it may avoid to deploy a second Abis link.
(2/5)
1
EGCH
2
EGCH
3
EGCH
4
EGCH
5
EGCH
6
EGCH
TRX
M-EGCH link
No more constraints in B9!
composed of
GCH Basic
GCH Extra
GCH Extra
GCH Extra
GCH Extra
GCH Basic
GCH Extra
GCH Extra
GCH Basic
1 to 5 GCHs depending on the TRX class (and only one GCH can use a basic Abis nibble)
11 |BSS Architecture | June 2007
1 to 36 GCHs (mixture of GCHs using basic, extra and bonus Abis nibbles)
All Rights Reserved Alcatel-Lucent 2006, #####
GCH Basic
Basic Abis nibbles and Extra Abis nibbles are statically mapped to a RTS. They can only be used in the EGCH of this RTS.
composed of
(3/5)
(4/5)
The basic Abis nibbles mapped to a RTS currently available for PS traffic (see Autonomous Packet Resource Allocation" feature to know the list of those RTSs) or mapped to a RTS used as MPDCH. The bonus basic Abis nibbles currently used for BCCH or static SDCCH channels:
The list of bonus Abis nibbles depends on the cell configuration.
(5/5)
The basic Abis nibbles mapped on a RTS allocated to MFS can be used in the MEGCH link of any TRX of the CELL. The extra Abis nibbles can be used in the M-EGCH link of any TRX of the BTS. The bonus Abis nibbles can be used in the M-EGCH link of any TRX of the BTS.
CS traffic
TRX 1 TRX 2
TCH TCH
GCH Basic
PS traffic
TRX 3
GCH Extra
Abis
GCH Extra GCH Bonus
TRX n
BTS
M-EGCH link n
n 12
GCH Basic
The Ater Resource Management in a given GPU is based on two complementary mechanisms:
GPU Ater TS margin, High Ater usage handling. A strong requirement is to ensure GPRS access in all the cells of the GPU (no cell shall be blocked due to an Ater congestion).
Management:
Release of some GCHs when the remaining number of free 64k Ater TSs in the GPU becomes too low (O&M parameter N_ATER_TS_MARGIN_GPU). For a given TRX, when releasing GCHs, it is ensured that:
Established_Nb_GCH remains higher than Min_Nb_GCH.
High Ater usage handling: Definition of the Ater usage of a GPU: Represents the consumption of Ater nibbles (by GCH channels) among the PCM links connected to the GPU, Ater usage can be either normal or high. Decision based on the comparison of the Ater nibble consumption with a threshold (Ater_Usage_Threshold O&M parameter).
Behaviour if Ater usage is high: Target_Nb_GCH values associated to TRXs of the GPU supporting some PS traffic will be reduced: GCH_RED_FACTOR_HIGH_ATER_USAGE O&M parameter. The reduction factor is only applied on PDCHs newly open. newly open PDCH means that no radio resources were previously allocated on this PDCH. When evaluating Target_Nb_GCH on a given TRX: If PDCH already open, no reduction is applied, If PDCH is newly open, reduction is applied.
Example:
Max_EGPRS_MCS = MCS-9, GCH_RED_FACTOR_HIGH_ATER_USAGE = 0,5 1) Ater usage = normal 2) Establishment of an EGPRS DL TBF on RTS0-3 Target_Nb_GCH = 4 * Nb_GCH(Max_EGPRS_MCS) = 4 * 4,49 = 18 3) Ater usage = high 4) Establishment of an EGPRS DL TBF on RST4-7 Target_Nb_GCH = 4 * Nb_GCH(Max_EGPRS_MCS) + 0,5 * 4 * Nb_GCH(Max_EGPRS_MCS) = 4 * 4,49 + 4 * 0,5 * 4,49 = 27 (< 36)
Goal: Avoid consuming transmission resources (Abis + Ater) in case of DL RLC data block retransmissions. Principles: Store for a certain time, in the memory of the TRE involved in the packet transfer mode with an MS, the DL RLC data blocks received from the RLC/MAC layer for this MS. Then, the RLC/MAC layer (in the MFS) can ask the TRE (in the BTS) to retransmit some data blocks. Data received from the MFS are stored in a buffer.
This mechanism is enabled / disabled depending on the EN_DL_RETRANS_BTS parameter, on the TRX HW capability and on the MFS-BTS round_Trip_delay:
HW generation of the TRE
EN_DL_RETRANS_BTS
Enabled Enabled disabled
Round_Trip_Delay
Network Architecture ASSESSMENT Architecture CHECK Parameter CHECK Dimensioning Methods RUN
Assessment of basic and bonus Abis nibbles from TRX and BTS configuration
TC
MT120
SMU TRCU TRCU TRCU
speech
MT120
Air
CS traffic PS traffic TRX 1 TRX 2 TRX 3
TCH TCH
A-bis
G C H B asic
SMU
BSC
Abis TSU Abis TSU Abis TSU Ater TSU Ater TSU Ater TSU
CS
CS
TRX n
M -EG C H link n
D yn a m ic A b is a llo c a tion
G C H B asic G C H B on u s G CH Extra
A-ter mux
CS+ PS PS
MFS
GPU board
DSP DSP DSP DSP
G CH Extra
Check the load and conectivity on DSP and GPU boards of MFS
B TS
data
SGSN
Calculate the needed radio TS for SDCCH, TCH and PDCH, then nb of TRXs
Calculate the needed extra Abis TS for real traffic on Abis Itf and the number of needed Abis links
GPU board
DSP DSP DSP DSP
Gb
Evaluate the required number of Gb 64K TS per GPU
Calculate the total number of Ater channels and the required number of Ater links
Unchanged methods
23 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Assessment of basic and bonus Abis nibbles from TRX and BTS configuration
SMU
TC
MT120
TRCU TRCU TRCU
speech
MT120
SMU TRCU TRCU TRCU
BSC
CS traffic PS traffic TRX 1 TRX 2
TCH TCH
CS
Ater TSU Ater TSU Ater TSU
G C H B asic
CS
Air
TRX 3 TRX n
M -EG C H link n
D yn a m ic A b is a llo c a tion
G C H B asic G C H B on u s G CH Extra
A-ter mux
CS+ PS PS
MFS
GPU board
DSP DSP DSP DSP
A-bis
G CH Extra
B TS
GPU board
SGSN
Calculate the needed extra Abis TS for real traffic on Abis Itf and the number of needed Abis links
Gb
data
EDGE activation (in case no EDGE in B8) Define Max. MCS Configurable Nb of Extra TS Same B8 EDGE approach : cf the example (slide 33 ) Set equal to a recommended initial value (cf slide 42 ) Best Effort approach : put as many Extra TS as there are available on the existing Abis links and respect to BSC limit
NW with GPRS CS3-4 and/or EDGE activation (in case no EDGE Assess the initial setting of EDGE in B8) Extra TS based on counter measurement. Fixed Nb of Extra TS Define Max. MCS needed corresponding Configurable Nb of Extra TS to TRX class (class 2 to 5) Same B8 EDGE approach A B8 B9 migration rule exists for the initialization of the B9 parameter :containing the Nb of extra Abis TS: NW w/o EDGE or with GPRS cf the example (slide 33 ) CS1-2 only Set equal to a N_EXTRA_ABIS_TS = 2*[(1*Nb-of-TRX-Trans-Pools-Type2)+(2*Nb-of-TRX-Trans-Pools-Type3)+(3*Nb-ofNot require Extra TS (Nb recommended initial value TRXTrans-Pools-Type4)+(4*Nb-of-TRX-Trans-Pools-Type5)] for each sector of the BTS of Extra TS = 0) (cf slide 42 ) in B9, is the sum of NB_EXTRA_ABIS_TS, In other words, N_EXTRA_ABIS_TS, for the BTSBest Effort approach : put as many Extra TS as there are for each sector of the BTS in B8. available on the existing B8: Not possible to Abis links and respect to optimize # Extra In order to have an iso-functionnality B8 B9 BSC limit it is recommended to configure MAX_GPRS_CS migration TS because of and MAX_EGPRS_MCS on the basis of the actual capability of the cell (depending on the TRX classes set static Abis In the cell) allocation B9: Extra Abis TS can be optimized, thanks to dynamic Abis allocation
27 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Design
Step III: Design BSC area/resource and Parenting board/port - Less BSC resource consumption expected in B9
Operational
Deploy Secondary Abis link (if needed) Activate GPRS (CS-3/4) / EDGE - B8 parameters - B9 parameters
Step I:
B8: Define TRX Class for each cells
Choose TRX Class (from class 1 to class 5) On how many TRXs are applied each TRX Class Example:
TRX Class Class 5 (Max CS-4 / Max MCS-9) Class 2 (Max CS-4 / Max MCS-5) Class 1 (Max CS-2) Number of TRX Class (per cell) 1 TRX 1 TRX 1 TRX Criterion For cells considered as "High" PS traffic For cells considered as "Medium" PS traffic For cells considered as "Low" PS traffic
Step II: B8: Analyse the Abis resource usage which mainly relates the fixed number of extra Abis TS statically mapping for each RTS.
Based on TRX Class, Number of TRX Class , TRX configuration of the BTS and Abis network topology (i.e.Chain, Star,or Ring). Is secondary Abis deployment needed?
B9: Design the Abis resource availability which can be dynamically allocated for extra Abis TS when needed (supporting PS traffic).
Based on defined Max. CS and MCS, TRX configuration of the BTS and Abis network topology (i.e.Chain, Star,or Ring). Nb of Extra Abis TS to be set at OMC per BTS Is secondary Abis deployment needed?
30 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Step III: B8 & B9: Design BSC resource/area and Parenting board/port
This step is common for B8 and B9. Design of BSC resource/area is usually needed as GPRS(CS-3,CS-4) / EDGE implementation consumes more BSC capacity even it should be less impact in B9. Parenting board/port to ensure that load is evenly spread among BSCs & Abis TSUs
B9 parameters
BTS
FR TRX #1 TRX_PREF_MARK = 1
Radio TS Configuration
TS0 TS1 TCH TS1 TCH / PDCH TS1 TCH TS1 TCH / PDCH TS1 TCH TS1 TCH / PDCH TS2 TCH TS2 TCH / PDCH TS2 TCH TS2 TCH / PDCH TS2 TCH TS2 TCH / PDCH TS3 TCH TS3 TCH / PDCH TS3 TCH TS3 TCH / PDCH TS3 TCH TS3 TCH / PDCH TS4 TCH TS4 TCH / PDCH TS4 TCH TS4 TCH / PDCH TS4 TCH TS4 TCH / PDCH TS5 TCH TS5 TCH / PDCH TS5 TCH TS5 TCH / PDCH TS5 TCH TS5 TCH / PDCH TS6 TCH TS6 TCH / PDCH TS6 TCH TS6 TCH / PDCH TS6 TCH TS6 TCH / PDCH TS7 TCH TS7 TCH / PDCH TS7 TCH TS7 TCH / PDCH TS7 TCH TS7 TCH / PDCH
BCCH
TS0
CELL #1
FR TRX #2 TRX_PREF_MARK = 0
SDCCH
TS0
CELL #2
FR TRX #1 TRX_PREF_MARK = 1
BCCH
TS0
FR TRX #2 TRX_PREF_MARK = 0
SDCCH
TS0
CELL #3
FR TRX #1 TRX_PREF_MARK = 1
BCCH
TS0
FR TRX #2 TRX_PREF_MARK = 0
SDCCH
Step II: Design Abis Resource (Max 32 Abis TS per 1 Abis link)
Calculation:
TS0 Basic Abis TS = 12 TS RSL/OML Abis TS Extra Abis TS = 1 TS = 2 Abis TS (per TRX) * 6 TRX (3 cells) = 2 TS = 24 TS
1 RTS of TRX Class 5 requires 4 extra Abis nibble (EGCH link) Total Extra Abis nibbles = 4 Abis nibbles * 8 RTS (per TRX) * 3 TRX Class 5 (3 cells) = 96 Abis nibbles So Total Extra Abis TS = 96 Abis nibbles / 4 Abis nibbles per TS = 24 Abis TS
B8 Design (cont):
Step III: Design BSC resource Check BSC resource consumption by this BTS
Based on Step II, there are..
12 TS used as Basic Abis TS 24 TS used as Extra Abis TS
So, Number of FR (equivalent) TRXs consuming the BSC capacity can be simulated as following: [12 (basic) + 24 (extra)] / 2 Abis TS equivalent to 1 TRX = 18 FR eq TRXs
B8 Operational:
2 Abis links needed to be implemented for this BTS
Primary Abis link
BSC
BTS
Step II: Design Abis Resource (Max 32 Abis TS per 1 Abis link)
Calculation:
TS0 Basic Abis TS TS RSL/OML Abis TS = 1 TS = 2 Abis TS (per TRX) * 6 TRX (3 cells) = 12 = 2 TS = 17 TS
Only TRX #2 supports GPRS/EDGE as TRX_PREF_MARK = 0. So, Max PDCH group size is 7 PDCHs with Max MCS = MCS-9. In B9 with M-EGCH link, MCS-9 requires 4.49 GCH 7 PDCH x 3 Cells x 4.49 GCH / PDCH = 94.3 GCH = 95 Abis nibbles 7 x 3 (Cells) Basic nibbles + 2 x 3 (Cells) Bonus nibbles = 27 (basic + bonus) nibbles (2 bonus nibbles / cell : BCCH in TRX#1+ SDCCH in TRX #2) 95 27 = 68 Extra Abis nibbles = 17 Extra Abis TS (4 Abis nibbles per TS)
37 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Calculation:
Only TRX #2 supports GPRS/EDGE as TRX_PREF_MARK = 0. So, Max PDCH group size is 7 PDCHs with Max MCS = MCS-9. In B9 with M-EGCH link, MCS-9 requires 4.49 GCH 7 PDCH x 3 Cells x 4.49 GCH / PDCH = 94.3 GCH = 95 Abis nibbles 7 x 3 (Cells) Basic nibbles + 2 x 3 (Cells) Bonus nibbles = 27 (basic + bonus) nibbles (2 bonus nibbles / cell : BCCH in TRX#1+ SDCCH in TRX #2) 95 27 = 68 Extra Abis nibbles = 17 Extra Abis TS (4 Abis nibbles per TS)
38 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
B9 Design (cont):
Max Total required Abis TS = 1 (TS0) + 12 (Basic) + 2 (RSL/OML) + 17 (Extra) Only 1 Abis link enough !!! = 32 Abis TS
Step III: Design BSC resource Check BSC resource consumption by this BTS
Based on Step II, there are..
12 TS used as Basic Abis TS Max 17 TS used as Extra Abis TS
So, Max number of (equivalent) FR TRXs consuming the BSC capacity can be simulated as following: [12 (basic) + Max 17 (extra)] / 2 Abis TS equivalent to 1 TRX = Max 14.5 FR eq TRXs
B9 Operational:
Only 1 Abis link needed to be implemented for this BTS
BSC
Primary Abis link
BTS
B9:
Only 1 Abis link enough !
Max 17 extra TS less extra TS,
thanks to M-EGCH statistical multiplexing & Dynamic Abis allocation features in B9.
This setting allows to support 2 MS MCS9 (4+1) in a typical tri-sector BTS having 21 basic nibbles (7 basic per cell) and 6 bonus nibbles: Needed nibbles = 2MS * 4PDCH * 4,49GCH_MCS9 = 36 nibbles Needed extra & bonus nibbles = Needed nibbles basic nibbles = 36 (2MS * 4PDCH ) = 28 nibbles Extra nibbles = Needed extra & bonus nibbles bonus nibbles = 28 6 = 22 nibbles
42 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
6 Extra TS
B9 Optimised phase: The first step is to assess the current Extra Abis TS whether it can handle the PS traffic without causing the QoS and performances degradation. If not, it will trigger the Abis re-dimensioning process to find the optimal Extra Abis TS. On Abis interface, Abis nibbles unavailability can lead to two different situations:
Service unavailability: TBF establishment / reallocation failure due to lack of enough Abis nibbles available for the establishment of the needed minimum number of GCH. Service degradation: Radio throughput reduction due to the fact that the number of available Abis nibbles is less than the needed one in order to support the targeted throughput In this case, RLC/MAC data blocks that cannot be sent in a given radio block will be delayed and sent in a future radio block
43 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
OR
If no Abis congestion has been detected looking at TBF establishment failure counters, The only way to check Abis dimensioning is applying the dimensioning methods described in next slides
45 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
CONs
Focused only on PS traffic Focused only on priority 2 Abis nibbles: Extra & Bonus nibbles
CONs
Complex to apply
ERLANG C
Measured _ extra & bonus _ Abis _ Traffic 1 Min(%TBF _ Abis _ Cong ,30% ) P 466 Measured _ extra & bonus _ Abis _ Traffic = 3600 %TBF _ Abis _ cong = Max(% DL _ TBF _ Abis _ cong ,%UL _ TBF _ Abis _ cong ) % DL _ TBF _ Abis _ cong = P105i x100% P91a + P91b + P91c + P91d + P91e + P91 f P105 j %UL _ TBF _ Abis _ cong = x100% P 62a + P62b + P62c P 438c
GoS= Grade Of Service
MCS Method
Nb Basic nibbles
Customers option
MCS Method is based on a traffic law (Knapsack model) allowing to dimension resources shared by different services (i.e. CS and PS) It is implemented thanks to 2G Traffic tool
using
Calculate GoS of all services in each cell
Traffic_ service_1
Traffic_ service_2
False
Nb of extra Abis TS
Nb of bonus nibbles
Stochastic method allowing to find a distribution f(n, prob(n)) in a multiservice environment, being n the number of used resources
No
No
Select minimumm between theoretically missing resources (according to the method given in slide 37 ) and the minimum between extra & bonus / MCS method results /4 N_EXTRA_ABIS_TS
51 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Round up
5 extra nibbles
6 extra nibbles
5 extra nibbles
Min
5 extra nibbles
See slide 37
All Rights Reserved Alcatel-Lucent 2006, #####
Is this a counter issue or is this due to a bad ressource management ? UNDER INVESTIGATION by TD
53 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
TC
MT120
SMU TRCU TRCU TRCU
speech
MT120
SMU TRCU TRCU TRCU
BSC
CS traffic PS traffic TRX 1 TRX 2
TCH TCH
G C H B asic
CS
CS
Air
TRX 3 TRX n
M -EG C H link n
D yn a m ic A b is a llo c a tion
G C H B asic G C H B on u s G CH Extra
A-ter mux
CS+ PS PS
MFS
GPU board
DSP DSP DSP DSP
A-bis
G CH Extra
B TS
SGSN
GPU board
DSP DSP DSP DSP
Gb
data
Calculate the total number of Ater channels and the required number of Ater links
(1/2)
(2/2)
BSC
(Number of GSL messages/sec per GPU)/ GSL_link_Max_capacity per GPU > 75%
Where:
Number of GSL messages/sec per GPU is calculated as a function of PS traffic (UL and DL TBF establishment requests) and of the number of cells mapped to the GPU/GP. Calculation details are provided in next slides GSL_link_Max_capacity per GPU is defined as: K_GSL * (1000ms/GSL_round_trip_delay) * number of configured GSL links per GPU/GP if GSL_round_trip_delay<500ms
Due to FR 3BKA20FBR202481
K_GSL * (1000ms/GSL_round_trip_delay) * (1/2) * number of configured GSL links per GPU/GP if GSL_round_trip_delay 500ms
Referecne GSL_round_trip_delay = 500ms (being K_GSL the maximum number of outstanding I frames on a GSL link) N.B. In B9MR4 the allowed range for K_GSL parameter has been enlarged in order to cope with satellite need: from a range of [0,16] to a range of [0,32])
59 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
(1/3)
Once a potential under-dimensioning has been detected on AterMux PS interface, AterMux PS dimensioning assessment method must be run at: I. BSC level (doing the hypothesis of a well balanced traffic distribution among the GPU/GP boards connected to the BSC)
AND II. GPU/GP level (in case of multi-GPU configuration and if no additional GPU/GP resource adding found through the method application at BSC level) in order to handle congestion situations due to unbalanced traffic/cell distribution/mapping on GPU/GP boards connected to the BSC. In this case:
a. a reshuffling operation should be done, before adding GPU/GP/Atermux resources, if needed, in order to be sure about the congestion root cause If the reshuffling does not solve the congestion situation, additional resources, according to the dimensioning method result, should be added N.B. If, running the dimensioning assessment method, more than 1 GPU/GP board are identified as underdimensioned (i.e. they are not able to handle the required traffic) the adding of GPU/GP boards will be done in an iterative way (1 GPU/GP at once) monitoring the consequent impact on the AterMux PS interface.
All Rights Reserved Alcatel-Lucent 2006, #####
b.
(2/3)
# AterMux PS
Aterlink GCH_Capacity
max
# GSL links
(at least 2 per GPU/GP)
GCH 0 16 32 60 88 116
# GPU/GP
All Rights Reserved Alcatel-Lucent 2006, #####
(3/3)
Aterlink GCH_Capacity
# AterMux PS
max
# AterMux PS per GPU (user traffic) # AterMux PS per GPU (estimated at BSC level)
# GSL links
(at least 2 per GPU/GP)
max
(1/3)
GPU/GP level
GPU1
#Needed GCH GPU1 = 200 #Needed GPU/GP = 1 #AterMux PS per GPU/GP = Max (200 / 112, 3) = 3
GPU2
#Needed GCH GPU2 = 300 #Needed GPU/GP = 1 #AterMux PS per GPU/GP = Max ( 300 / 112, 3) = 3
1) Reshuffle operation is done in order to try to balance traffic between the two GPUs 2) Add 1 AterMux PS links on both GPUs.
(2/3)
GPU/GP level
GPU1
#Needed GCH GPU1 = 160 #Needed GPU/GP = 1 #AterMux PS per GPU/GP = Max (160 / 112, 2) = 2
GPU2
#Needed GCH GPU2 = 240 #Needed GPU/GP = 1 #AterMux PS per GPU/GP = Max (240 / 112, 2) = 3
1) Reshuffle operation is done in order to try to balance traffic between the two GPUs 2) If the reshuffle operation has not the wanted effect, add 1 AterMux PS to GPU2.
64 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
(3/3)
GPU/GP level
GPU1
#Needed GCH GPU1 = 200 #Needed GPU/GP = 1 #AterMux PS per GPU/GP = Max (200 / 112, 3) = 3
GPU2
#Needed GCH GPU2 = 300 #Needed GPU/GP = 2
1) Reshuffle operation is done in order to try to balance traffic between the two GPUs 2) If Reshuffle has not the wanted effect:
#Needed GCH at BSC = 500 #Needed GPU/GP = 3 #AterMux PS per BSC = 500 / 112 = 5 #AterMux PS per GPU/GP = 5 / 3 = 2
65 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
If GSL load > 80% ( slide 58) than add 1 GSL link on the GPU/GP on which the GSL overload is observed
2
(for security reason)
max
This method was applied and validated in case of Satellite Ater links Here an extension to Terrestrial links is presented but no field feedback exists on that
# GSL links
N.B. In case of GSL under-dimensioning due to a limitation in terms of number of GSL Messages per second treatement capability, an alternative solution to GSL link adding is the tuning of K_GSL parameter (WARNING: no field feedback available on this practise)
66 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
1) Msg on GSL due to RAE4 mechanism 2) Msg on GSL due to PS traffic: 2.1) Msg on GSL due to PS/CS paging
1 x (Nb_PS_paging/sec+ Nb_CS_paging/sec)
2.2) Msg on GSL due to PS data traffic (TBF requests): TBF (UL & DL) corresponding to RA update UL TBF without sig on GSL (eg. ACK of FTP DL transfer) TBF (UL & DL) corresponding to PS data (3 cases) Low GSL usage (eg. Ping 5 sec) Medium GSL usage High GSL usage (worst case) 2.5 3.5 5 x Nb_TBF_Data_req/sec 1.7 x Nb_TBF_Sig_req/sec 0 x Nb_UL_TBF_req_noGSL/sec
Nb GSL messages/sec
67 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
PS traffic If #TBF req / sec < 20 TBF/sec Nb of Msg on GSL (TBF req) High Available GCH If ater congestion observed Low 2.5 3.5 High 3.5 Low 5
The values provided in the table above are relibale if T_GCH_INACTIVITY=3sec and T_GCH_INACTIVITY_LAST=20sec (default setting). Otherwise (I.e. T_GCH_INACTIVITY=2sec and T_GCH_INACTIVITY_LAST=6sec GSL load will be greater due to the consequent increase of deallocation / allocation messages).
START
QoS acceptable ?*
(i.e. UL TBF estab success rate >80%)
No
OR
Select hours with acceptable QoS * (i.e. for 90% of cells) Compare PS traffic in the selected hours with traffic observed on a similar BSC (reference BSC) Estimate PS traffic at busy hours on the basis of the reference BSC (through a simple proportion based on the respective number of cells)
# GSL links
70 |BSS Architecture | June 2007
75% x GSL_Link_Max_Capacity (see slide 59 for computation If the method is applied at BSC level, the total capacity (for all GPU/GP) must be kept
All Rights Reserved Alcatel-Lucent 2006, #####
With quantile=99,9%
N.B. The dimensioning assessement of AterMux interface can be done both at BSC and at GPU/GP level (i.e. in case of multi-GP(U) configuration)
71 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_Traffic estimation
Required_GCH_Traffic estimation
Two different methods can be used for the estimation of the Required_GCH_traffic quantity:
Method 1: driven by the estimation of the required traffic as a function of the measured GCH traffic and of Ater/GPU congestion P383a, P384, P201, P202, P402 Congestion
Stable Network
Method 2: driven by the estimation of the required traffic as a function of the measured GCH and radio PS traffic
Required_GCH_Traffic estimation
Method 1
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Required_GCH _traffic
350
300
250
200
150
100
50
0 06/03/2007 : 00h 04h 08h 12h 16h 20h 07/03/2007 : 00h 04h 08h 12h 16h 20h 08/03/2007 : 00h 04h 08h 12h 16h 20h 09/03/2007 : 00h 04h 08h 12h 16h 20h 10/03/2007 : 00h 04h 08h 12h 16h 20h 11/03/2007 : 00h 04h 08h 12h 16h 20h 12/03/2007 : 00h 04h 08h 12h 16h 20h 13/03/2007 : 00h 04h 08h 12h 16h 20h 14/03/2007 : 00h 04h 08h 12h 16h 20h 15/03/2007 : 00h 04h 08h 12h 16h 20h
Required_GCH_Traffic estimation
Method 1
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Required_GCH _traffic
350
300
250
P100c Measured_GCH 3600 _traffic Congestion = Max{Ater _ Congestion, GPU _ Limitation} = Max{% Ater _ Cong ,% DSP _ Cong ,% DSP _ load ,%CPU _ overload } Measured _ GCH _ Traffic = P383a ; 3600 P384 % DSP _ Cong = ; 3600 Max( P 201, P 202) % DSP _ load = ; 3600 P 402 %CPU _ overload = 3600 % Ater _ Cong =
200
150
100
50
Since Method 1 is reliable if congestion is less than 30% and it does not take into account High Ater usage condition, a complementary method, Method 2, has been defined
74 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
06/03/2007 : 00h 04h 08h 12h 16h 20h 07/03/2007 : 00h 04h 08h 12h 16h 20h 08/03/2007 : 00h 04h 08h 12h 16h 20h 09/03/2007 : 00h 04h 08h 12h 16h 20h 10/03/2007 : 00h 04h 08h 12h 16h 20h 11/03/2007 : 00h 04h 08h 12h 16h 20h 12/03/2007 : 00h 04h 08h 12h 16h 20h 13/03/2007 : 00h 04h 08h 12h 16h 20h 14/03/2007 : 00h 04h 08h 12h 16h 20h 15/03/2007 : 00h 04h 08h 12h 16h 20h
Required_GCH_Traffic estimation
Method 2 METHOD 2:
448 y = 5,3905x + 21,057
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Required_GCH_Traffic
336
224
Resources saturation
112
0 0 10 20 30 40 50 60 70 80
Measured GCH traffic = Function_of(PDCH traffic, traffic profile, EGPRS penetration, CS3/CS4 activation, Ater & Abis congestion, high ater usage, available resources, parameter configuration (i.e. MAX_GPRS_CS, MAX_EGPRS_MCS, T_GCH_INACTIVITY_LAST, EN_FAST_INITIAL_GPRS_ACCESS,EN_EGPRS, etc)) Required_GCH_traffic has a quasi-linear relashionhip with the PDCH traffic (observed on several networks) if no congestion nor strong GCH usage reduction due to High Ater usage present
75 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_Traffic estimation
An example
Following the 4th link adding
(1/3)
Stable Network
Needed_GCH
560
ERLANG C
336
224
112
0 0 10 20 30 40 50 60 70 80
N.B. If the PDCH traffic reaches values around 70erl (as previously observed) the related estimated GCH traffic will require the adding of a 5th link
76 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_Traffic estimation
An example
4 links: from 12/03 to 27/03
gch traffic
448
(2/3)
Stable Network
y = 5,3905x + 21,057
336
224
3 links
112
0 0 10 20 30 40 50 60 70 80
Pdch traffic
Abnormal resource management: in the meantime UL TBF Establishment Failure due to BSS Fail equal to 25%
Required_GCH_Traffic estimation
An example
(3/3)
Stable Network
gch traffic
448
4 links: 16 more cells (EDGE/CS3/CS4 activated on 3 cells over 16) EN_FAST_INITIAL_GPRS_ACCESS enabled
y = 5,3905x + 21,057
336
224
112
0 0 10 20 30 40 50 60 70 80 90
Pdch traffic
EN_FAST_INITIAL_GPRS_ACCESS = enable represented a workaround for this abnormal resource management (see slide 112 for related recommandations)
78 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_Traffic estimation
EGDE, CS3-CS4 activation
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
With the activation of EDGE and/or the use of GPRS higher coding schemes (CS3/CS4) the consumption of AterMux transmission resources (GCH) per radio resource (PDCH) increase. The increase factor will be a function of: the transition type the target service penetration (i.e. %EDGE with respect to GPRS) the traffic profile
Required_GCH_Traffic estimation
Stable Network
Yes
Increase_factor estimated on the basis of the Avg_target_nb_GCH_per_PDCH (depending on target service penetration)
Execute transition
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
% of Radio Resources (PDCH) supporting at least one TBF established in EGPRS mode on a cell with MAX_EGPRS = MCSx %_PDCH_EGPRS % of Radio Resources (PDCH) supporting only TBF established in GPRS mode on a cell with MAX_GPRS = Csy %_ PDCH_GPRS Transition type a b c d e Increase_Factor
[(%_PDCH_EGPRS*4,49)+(%_PDCH_GPRS *1,64)]/1 [1,64]/1 [(%_PDCH_EGPRS*4,49)+(%_PDCH_GPRS *1)]/1 [(%_PDCH_EGPRS*4,49)+(%_PDCH_GPRS *1,64)]/ 1,64 (%_PDCH_EGPRS*4,49)+(%_PDCH_GPRS* 1,64)/(%_PDCH_EGPRS*4,49)+(%_PDCH_ GPRS*1)
CS1/CS2
b a c
CS3/CS4
d
CS3/CS4 & EDGE
EDGE
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Increase Factor
Required_GCH_traffic
After EDGE/CS3/CS4
GCH traffic
Y=a2x + b2
Y=a1x + b1
Before EDGE/CS3/CS4
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
EDGE / CS3-CS4
336
GCH traffic
224
CS1-CS2
y = 1,9111x + 20,368
PDCH traffic
0 0 10 20 30 40 50 60 70 80
Observed increase factor = 5,309 / 1,9111 = 2,8 This difference can be explained by the fact that: the estimated value is based on average PDCH_GPRS/PDCH_EGPRS values the number of established_GCH for a given number of PDCH X is actually the rounded up value of (X*avg_target_nb_GCH_per_PDCH) the final and initial trends are estimated trends
All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Being the increase_factor for this kind of transition (b) equal to 1,64 (cf. previous slide):
224
168
ERLANG C
y = 1,798x + 7,6258 56
0 0 5 10 15 20 25 30 35
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
GCH traffic
168
CS3-CS4
Forecasted trend
BSC level
y = 1,798x + 7,6258 56
CS1-CS2
0 0 5 10 15 20 25 30 35
PDCH traffic
The forecasted trend is very close to the observed one after CS3-CS4 activation Additional AterMux PS resources are required (as previously recommended)
Required_GCH_Traffic estimation
High level recommendation
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Taking into account the method described in previous slides, the following high level recommandations are provided in order to anticipate EDGE / CS3-CS4 activation: CS1-CS2 CS1-CS2 CS3-CS4 CS3-CS4 and EN_EGPRS=disabled CS3-CS4 and EN_EGPRS=enable (MAX_EGPRS_MCS=9) CS3-CS4 and EN_EGPRS=enable (MAX_EGPRS_MCS=9)
maximum_number_of_GCH represents the maximum observed value of P100b (in B8) or P100f (in B9). If P100b / P100f counter value is not available maximum_number_of_GCH=112 (capacity of 1 PS link)
Required_GCH_traffic estimation
BTS moving
BTS Moving
BTS TRX BTS TRX BTS TRX BSC
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Method 1
Cell Zone related To BTS to move Cell Zone related To BSC before BTS moving
GCH traffic
Method 2
Required_GCH_traffic
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
+
Required_GCH_traffic
87 |BSS Architecture | June 2007
PDCH traffic Cell Zone related PDCH contribution To BSC before BTS moving due to Cell Zone Related to BTS to move
All Rights Reserved Alcatel-Lucent 2006, #####
Required_GCH_traffic estimation
An example (I)
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
In the following graph the evolution of the relashionship between GCH traffic and PDCH traffic in a BSC, following the activation of EDGE and the moving of some BTS, is put in evidence:
560 y = 5,7292x + 10,705
448
(4),(2)
Karl-D: no edge with 2 links 336 gch usage y = 5,1084x + 12,146
(3)
Karl-D: edge activated with 2 links Karl-D: edge activated with 4 links karl-C: no edge Karl-C contrib Karl_D+Karl_C (6 links w/ 2 GPU)
224
(2) (1)
112 y = 1,9907x + 13,44
Karl-D+Karl-C trend
(1)
y = 1,8171x + 8,2315 0 0 20 40 60 pdch usage 80 100
(3) Cell Zone related Cell Zone related To BSC before BTS moving To BTS to move
Why is a resource saturation observed even if the number of AterMux links is the one 120 recommended looking at BSC traffic level ?
In this example: BSC before BTS moving is called Karl-D Cell zone related to BTS to move is called Karl-C (xx) gives indication about operation sequence
Required_GCH_traffic estimation
An example (II)
GPU 1_10
336
Required_GCH_Traffic estimation
Stable Network Feature Traffic activation evolution
112
In case of Multi-GPU Configuration, even if the total Number of aterMux links at BSC level would be enough for the Required GCH traffic, if the Traffic is not balanced between The two GPUs, additional link(s) Per GPU could be needed
336
0 0 5 10 15 20 25 30 35 40
224
112
10
20
30
40
50
60
70
GPU/GP board
Packet Management Unit (PMU): PDCH management PS Paging management Gb stack management SGSN
PPC
BSC
DSP
PMU
PTU
PPC/CPU DSP Power Limitation Memory Limitation Power Limitation Memory Limitation P76a P392a P201 (thr_1 ) Dimensioning P77a P384 P392b P202 (thr_2 ) indicators => P402 (thr ) QoS P105e UL TBF estab P203 P105c indicators P105f BSS Failure P204 P105d (TBF establ)
CPU Cong
BSS Fail
DSP Load
DSP Cong
In Alc_Mono_GPRS_Telecom RNO report MFS parameters: Thr = PMU_CPU_Overload (Default=95%) Thr_1 = DSP_Load_Thr_1 (Default=85%) Thr_2 = DSP_Load_Thr_2 (Default=95%)
In case of B8
B9 migration
OR
B9 only
Memory limitation
This kind of limitation should be no more observable starting from B9MR1ed6QD2 AND
P392b = 1000 (B9MR1 & B9MR4 for legacy MFS) P392b = 4000 (B9MR4 for Mx MFS)
+
Workaround on issue no more observed from B9MR1ed6QD2
Number of GPU
Being: GPU_GCH_Capacity the maximum number of GCHs that the GPU can handle, Needed GCHs the number of GCH resources to be handled (for the computation of this quantity refer to slides dedicated to aterMux dimensioning ) GPU_for_Ms_context_handling a quantity equal to 1 if a GPU memory limitation due to a too big number of MS contexts is observed (issue no more observed from B9MR1ed6QD2) and no additional GPU/GP boards needed because of GPU GCH capacity limitation GPU_For_Power_Limitation a quantity equal to 1 if a GPU power limitation is observed, no additional GPU/GP boards needed because of GPU GCH capacity limitation and GPU_for_Ms_context_handling equal to 0.
94 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
+
Workaround on issue no more observed from B9MR1ed6QD2
Number of GPU
IfBeing: additional GPU/GP resources are needed, the following operational recommandations apply: GPU_GCH_Capacity the maximum number of GCHs that the GPU can handle, 1. Needed assessment has been done at BSC be handled (for the computation of this If the GCHs the number of GCH resources to level, the required additional resources will be added dedicated to aterMux dimensioning quantity refer to slides in one shot operation ) 2. GPU_for_Ms_context_handling done at GPU/GP level,GPU memory limitation due to a If the assessment has been a quantity equal to 1 if a the adding of additional too big number of a contexts is will be (issue no more observed from B9MR1ed6QD2), GPU/GP boards, for MS given BSC,observed done in an iterative way (1 GPU/GP at once), no additional GPU/GP boards needed because of GPU PS and GPU/GP dimensioning monitoring the consequent impact on the AterMux GCH capacity limitation
GPU_For_Power_Limitation a quantity equal to 1 if a GPU power limitation is observed, no additional GPU/GP boards needed because of GPU GCH capacity limitation and GPU_for_Ms_context_handling equal to 0.
GPU_GCH_Capacity calculation
(1/3)
In order to find the GPU_GCH_Capacity, the following limitations must be taken into account:
Maximum number of GCH per GPU is:
B9MR1: 480 N_ATER_TS_MARGIN_GPU*4 B9MR4: 480 N_ATER_TS_MARGIN_GPU*4 (for legacy MFS) 1560 N_ATER_TS_MARGIN_GPU*4 (for Mx MFS)
Maximum number of PDCH per GPU is dynamic depending on the used coding schemes (for details refer to Architecture section describing GPU/GP capacity in terms of PDCH) Maximum number of GCH is also dynamic because the number of GCH per PDCH depends on the used coding scheme. Therefore the maximum number of GCHs that the GPU/GP will be able to handle can be obtained knowing the:
(M)CS distribution of the analysed network (P55x & P57y counters) The maximum number of PDCH per coding scheme The maximum number of GCH per PDCH per coding scheme
GPU_GCH_Capacity calculation
Reminders
(2/3)
GPU_GCH_Capacity calculation
(3/3)
In DL direction the maximum number of GCHs that a GPU/GP will be able to handle is defined as:
Max_DL_GCH_GPU = (%CS1 * max_PDCH_CS1 * max_DL_GCH_per_PDCH_CS1)+ (on all
coding schemes)
In UL direction the maximum number of GCHs that a GPU/GP will be able to handle is defined as:
Max_UL_GCH_GPU = (%CS1 * max_PDCH_CS1 * max_UL_GCH_per_PDCH_CS1)+ (on all
coding schemes)
GCH_GPU_Capacity
35
30
336 25
20 224 15
10 112
0 06/03/2007 : 00h 07/03/2007 : 00h 08/03/2007 : 00h 09/03/2007 : 00h 10/03/2007 : 00h 11/03/2007 : 00h 12/03/2007 : 00h 13/03/2007 : 00h 14/03/2007 : 00h 15/03/2007 : 00h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h 06h 12h 18h
GPU is not able to support the required GCH traffic it is recommended to add a new GPU or to use a GP board instead of the needed GPUs boards N.B. If additional resources are added on AterMux the congestion at GPU level will become bigger (bottleneck moving from AterMux interface to GPU)
99 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
GPU_for_MS_context_handling
What does it mean ?
In B9 GPU board experiences a memory limitation that is quite more restrictive than in B8 regarding the maximum number of MS contexts (active or idle): B8 limitation is fixed at 2500 MS contexts B9MR1 limitation is fixed at 1000 MS contexts B9MR4 limitation is fixed at 1000 MS contexts for legacy MFS and at 4000 for MxMFS The consequence of the limit reaching observed in B9 versions BEFORE B9MR1ed6QD2 was an abnormal increase of UL TBF establishment failures due to BSS problem The workaround proposed up to now for handling this problem is the adding of 1 more GPU (GPU_for_MS_context_handling=1)
GPU_for_MS_context_handling
An example
1200
60,0%
1000
50,0%
800
40,0%
600
30,0%
400
20,0%
200
10,0%
0 03h 06h 09h 12h 15h 18h 21h 03h 06h 09h 12h 15h 18h 21h 03h 06h 09h 12h 15h 18h 21h 03h 06h 09h 12h 15h 18h 21h 03h 06h 09h 12h 15h 18h 21/10/2006 : 00h 22/10/2006 : 00h 23/10/2006 : 00h 24/10/2006 : 00h
0,0%
25/10/2006 : 00h
21h
GPU_for_MS_context_handling calculation
Retrieve indicators for 5 working days
NO
GPU_for_MS_context_handling=0
YES
P392bBSC = 1000 when BSS_fail_ratemax Observed for all days with at least two occurrences of P392bBSC = 1000
NO
GPU_for_MS_context_handling=0
YES
Observed QoS acceptable for the customer ?
NO
GPU_for_MS_context_handling=1
YES
GPU_for_MS_context_handling=0
102 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
An intra-cell pre-emption mechanism allows (being the Nb of MS context equal to 1000) to reduce impacts of MS Context limit reaching on QoS, improving per cell memory usage at GPU level
GPU Memory for MS context Cell A
1. New MS context needed in Cell A: NOK 2. New MS context needed in Cell B: OK
Cell B
Cell B
An inter-cell pre-emption mechanism allows (being the Nb of MS context equal to 1000 or 4000) to further reduce impacts of MS Context limit reaching on QoS, improving memory usage at GPU level
GPU Memory for MS context
1. New MS context needed in Cell A: OK 2. New MS context needed in Cell B: OK
Idle MS context Active MS context Intra and Inter cell pre-emption mechanisms could lead to an increase of PMU CPU load in GPU/GP
104 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
GPU_for_Power_Limitation
PMU CPU load
No field feedback exists, up to now, on this method
In case of any other abnormal event / behavior observed during the CPU loaded hours a specific analysis should be done in order to identify the actual problem root cause
NO
GPU_for_Power_Limitation=0
YES
P402/3600 >0,1% and (max(P105e/P105f) > 1% OR GPU reboots observed during the CPU loaded hours)
NO
GPU_for_Power_Limitation=0
YES
GPU_for_Power_Limitation=1
WARNING!: In case of B8 B9 migration, an additional condition must be checked for GPU_for_Power_Limitation definition: If GPU (not GP !) and P76a>40% in B8 then GPU_for_Power_Limitation=1
105 |BSS Architecture | June 2007 All Rights Reserved Alcatel-Lucent 2006, #####
GPU_for_Power_Limitation
DSP CPU load
No field feedback exists, up to now, on this method
In case of any other abnormal event / behavior observed during the CPU loaded hours a specific analysis should be done in order to identify the actual problem root cause
NO
GPU_for_Power_Limitation=0
YES
Max(P201,P202)/3600 >0,1% and (max(P203/P204) > 1% OR GPU reboots Observed during the CPU loaded hours)
NO
GPU_for_Power_Limitation=0
YES
GPU_for_Power_Limitation=1
TC
MT120
SMU TRCU TRCU TRCU
speech
MT120
SMU TRCU TRCU TRCU
BSC
CS traffic PS traffic TRX 1 TRX 2
TCH TCH
G C H B asic
CS
CS
Air
TRX 3 TRX n
M -EG C H link n
D yn a m ic A b is a llo c a tion
G C H B asic G C H B on u s G CH Extra
A-ter mux
CS+ PS PS
MFS
GPU board
DSP DSP DSP DSP
A-bis
G CH Extra
B TS
SGSN
GPU board
DSP DSP DSP DSP
Gb
data
Evaluate the required number of Gb TS ( E1 links) per GPU
Gb interface dimensioning
It is recommended to configure 1 NSVC (and one PVC) per E1 link in order to better take advantage of the available global bandwidth It is recommended to have at least 2 E1 links for redundancy reason The needed number of TS per NSVC (PVC) can be estimated through the following method:
Required_Gb_Traffic
ERLANG C
Required number of Gb TS
With quantile=99,9%
Re quired _ Gb _ Traffic = Measured _ Gb _ Traffic + 15% M arg in Measured _ Gb _ Traffic = Gb _ resources _ busy _ time data _ volume 1 Max( P 45, P 46) 1 = = 3600 Gb _ TS _ bandwith 3600 [64kbit / s ] / 8 3600
Where: - P45 (respectively P46) is the number of kilo bytes received from the SGSN (respectively sent to the SGSN) at PVC level
Network Architecture ASSESSMENT Architecture CHECK Parameter CHECK Dimensioning Methods RUN
3. An alternative solution to EN_FAST_INITIAL_GPRS_ACCESS = enable, allowing as much as possible GCH resource wasting, and being more traffic driven is the tuning of T_GCH_INACTIVITY_LAST and T_GCH_INACTIVITY to their default value (Recommended option)
PROs QoS could be improved due to less resource allocation / deallocation processing GCH maintained only where needed
112 |BSS Architecture | June 2007
CIR represents the Committed Information Rate (in Kbits/sec) at Frame Relay level for UL direction (from MFS to SGSN) NIR represents the Normal Information Rate (in Kbits/sec) at Frame Relay level for UL direction (from MFS to SGSN) CBS represents the Committed Burst Size (in Kbytes) during the time interval T EBS represents the Excess Burst Size (in Kbytes) during the time interval T EIR represents the Excess Information Rate (in Kbits/sec) at Frame Relay level for UL direction (from MFS to SGSN) ACCESS_RATE_BC represents the maximum capacity of the Gb link (number of Gb TS * 64 Kbit/sec)
CBS T time
CIR = (CBS * 8) / T EIR = [(EBS + CBS)*8] / T CIR <= NIR <= EIR <= ACCESS_RATE_BC
Warning: the default value for CIR / NIR is 0 Update this parameter if no direct MFS-SGSN connection
PS dimensioning
Parameters (I)
HMI name
Definition
Subsystem
Instanc e
OMC-R access
Type
Def value
Range
Unit
MAX_EGPRS_MCS
Maximum Modulation and Coding Scheme used for EGPRS traffic in the cell. Maximum coding scheme used for GPRS traffic in the cell.
MFS
Cell
Changeable
Number
[1,9]
None
MAX_GPRS_CS
BSC
Cell
Changeable
Number
[1,4]
None
EN_EGPRS
BSC
Cell
Changeable
Flag
[0,1]
None
Number of extra Abis (64k) timeslots configured for a BTS. Minimum number of master and slave PDCHs that are always allocated to the MFS. Maximum number of slave and master PDCHs that can be allocated to the MFS in the cell. Maximum number of slave and master PDCHs that can be allocated to the MFS when the CS traffic is high. This flag indicates whether or not one Slave PDCH is available for (E)GPRS traffic in the cell (fast initial PS access feature). - In a Non Evolium BTS, that corresponds to an established PDCH being able to support incoming (E)GPRS traffic. - In an Evolium BTS, that corresponds to a PDCH being able to support incoming (E)GPRS traffic, that PDCH being located on an established TRX (i.e. the TRX owns an M-EGCH link) Two definitions are possible : - If EN_FAST_INITIAL_GPRS_ACCESS = enabled : number of GCHs required to be established due to the Fast Initial PS Access feature, - If EN_FAST_INITIAL_GPRS_ACCESS = "disabled : number of GCHs to keep established when there is no more (E)GPRS traffic in a cell (while the T_GCH_INACTIVITY_LAST timer is running).
BSC MFS
BTS Cell
Changeable Changeable
Number
0 0
[0,60]
None
Number MFS Cell Changeable Number MFS Cell Changeable Number MFS MFS Changeable Flag 0 0 0
[0,127]
None
MAX_PDCH
[0,127]
None
MAX_PDCH_HIGH_LOAD
[0,127] [0,1]
None None
NEW : EN_FAST_INITIAL_GPRS_ACCESS
NEW : N_GCH_FAST_PS_ACCESS
MFS
MFS
None (DLS)
Number
[1,5]
None
PS dimensioning
Parameters (II)
HMI name
Definition
Subsystem
Instanc e
OMC-R access
Type
Def value
Range
Unit
NEW : T_GCH_Inactivity
- For Non Evolium BTS : Timer to postpone the release of one slave PDCH, when it does not support any (E)GPRS traffic. - For Evolium BTS : Timer to postpone the release of the "unused" GCHs of the M-EGCH link of a TRX (the condition for some GCHs of the M-EGCH link of a TRX to become "unused" is that some TBFs established on that TRX were released). - For Non Evolium BTS : Timer to postpone the release of the last established slave PDCH of a cell, when it does not support GPRS traffic anymore. - For Evolium BTS : Timer to postpone the release of the last N_GCH_FAST_PS_ACCESS GCHs established in a cell, when the last TBF has been released in the cell. Default value of the (E)GPRS multislot class assumed at TBF establishment when the actual MS (E)GPRS multislot class is unknown. - In an Evolium BTS : The default MS class is used as an input of the best candidate TBF allocation computation process when the MS class is not known on UL TBF establishment - In a Non Evolium BTS : The default MS class is used as an input of the best candidate TBF allocation computation process when the MS class is not known on UL TBF establishment. In the PDCH anticipation process on UL TBF establishment : the default MS class is used to determine how many PDCHs are established in advance to anticipate the concurrent DL TBF establishment.
MFS
BSS
Changeable
Timer
[1,100]
sec
NEW : T_GCH_Inactivity_Last
MFS
BSS
Changeable
Timer
20
[1,200]
sec
GPRS_MULTISLOT_CLASS_DEF_V ALUE
MFS
BSS
Changeable
Number
[1,2,4,8]
None
Ater_Usage_Threshold
Threshold (percentage of used Ater nibbles, in a GPU) above which the Ater usage is said high.
MFS
BSS
Changeable
Number
70
[1,100]
PS Dimensioning
Parameters (III)
HMI name
NEW : GCH_RED_FACTOR_HIGH_ATER _USAGE NEW : N_ATER_TS_MARGIN_GPU
Definition
Reduction factor of the number of GCHs targeted per PDCH, when the Ater usage is high.
Subsystem
MFS
Instanc e
cell
OMC-R access
Changeable
Type
Number
Def value
0.75
Range
[0,1,1]
Unit
none
Number of free 64k Ater TSs that are kept in reserve in order to be able to serve some prioritary requests in cells managed by the GPU. The prioritary requests are the GCH establishment requests launched when the first TBF has to be established in a cell. Note : In case of non-Evolium BTS, those are PDCHs that will be established instead of GCHs. Average bitrate per PDCH for non-Edge capable terminals in this cell
MFS
BSS
Changeable
Number
[0,10]
none
R_AVERAGE_GPRS
MFS
cell
Changeable
Number
12000
[0,20000]
bit/s
R_AVERAGE_EGPRS
Average bitrate per PDCH for Edge capable terminals in this cell
MFS
cell
Changeable
Number
30000
[0,59200]
bit/s
PMU_CPU_Overload
Threshold above which a GPU enters the PMU CPU overload state.
MFS
GPU
None (DLS)
Number
95
[0,100]
NEW: DSP_LOAD_THR_1
Threshold beyond which a given DSP is considered as "loaded" in terms of CPU load by RRM.
MFS
MFS
None (DLS)
Threshol d
85
[1,95]
NEW: DSP_LOAD_THR_2
Threshold beyond which a given DSP is considered as overloaded in terms of CPU load by RRM.
MFS
MFS
None (DLS)
Threshol d
95
[1,95]
PS Dimensioning
Parameters (IV)
HMI name
K_GSL (MFS)
Definition
Maximum number of outstanding I frames on a GSL link.
Subsystem
MFS
Instanc e
MFS
OMC-R access
None (DLS)
Type
Number 7
Def value
Range
[1,16] in B9MR1 [1,32] in B9MR4
Unit
none
MFS
BC
Displayed
Number
64
[64,1984]
Kbit/s
CBS
MFS
PVC
Changeable
Number
[0,248]
Kbytes
EBS
MFS
PVC
Changeable
Number
None
[0,248]
Kbytes
NIR
MFS
PVC
Changeable
Number
[0,1984]
Kbit/s
CIR
MFS
PVC
Changeable
Number
[0,1984]
Kbit/s
For ARCHITECTURE CHECK: AMT .NET AMT v5 script OMC AMT .NET excel export files
RNL & EML files For DIMENSIONING METHODS RUN: 2G Traffic Tool (B9 only)
TSL script 13 .txt files
NPA / OMC
2G Tool
Excel templates
RNO BestOf or manual search
Air Interface
Gb dimensioning
Power limitation and MS context Input data Input data limitation imported but imported but manual manual analysis analysis needed needed X X
Documentation
B9: BSS Architecture Service Guidelines (ed2 Released available and ed3 available for review). B9 Seminar(s) presentation(s) Quick & Light dimensioning recommendations for B9 release Complete list of needed indicators for dimensioning assessment Guidelines on how to identify linked BSC-GPU-PVC-LAPD objects
Tools
AMT .NET : Reference to official AMT .NET web page (version 2.0 delivered in w22) 2G Traffic tool + data check tool: user guide, executable and TSL script (with related execution process) for traffic indicator values export (version 1.2.0.2 delivered in w23). Excel templates for dimensioning assessment
(*) http://aro.tm.alcatel.ro
122 |BSS Architecture | June 2007
Technology 2G
B9 Release
Questions
& Answers
www.alcatel-lucent.com