Transcript Capacity and Load Sharing in Dual

Capacity and Load Sharing in Dual-Mode Mobile Networks
Author: Juha Peura
Supervisor: Prof. Patric Östergård
Instructor: M.Sc. Jarkko Laari, DNA Finland Oy
Agenda
• Background
• Objectives of the thesis
• Capacity
• Load sharing
• Conclusions
2
Background
• The amount packet data in mobile networks
have increased dramatically
• Operators have to guarantee quality of service
• New solutions needed for traffic handling
100 %
UMTS Downlink Data [MB]
90 %
80 %
60 %
50 %
40 %
30 %
20 %
10 %
24.2.2008
10.2.2008
27.1.2008
13.1.2008
30.12.2007
16.12.2007
2.12.2007
18.11.2007
4.11.2007
21.10.2007
7.10.2007
23.9.2007
9.9.2007
26.8.2007
12.8.2007
0%
29.7.2007
Percentage
70 %
Date
3
Objectives of the thesis
• What are the main performance bottlenecks in
todays mobile networks?
• Is it possible to ease the situation with load sharing
algorithms?
4
Capacity – channel elements
• Channel element is a measure of node B hardware resources
• Separate CE pools for UL/DL, common to all sectors
• One 12.2 kbps speech service uses one channel element
• HSUPA takes up to 32 CEs, non-serving cells
reserves also CEs
• CEs are a capacity bottleneck in uplink direction
• Many RAX-boards have 64 CEs
• Suggested minimum is 128 CEs
5
Capacity – Iub transmission
• Symmetric bit pipe between base station and RNC
• 3GPP have specified two transport methods: ATM and IP
• Implemented using so called E1s
• Maximum throughput of one E1 is 2 Mbps
• For speech traffic one E1 has been sufficient
• Fast packet connections need multiple E1s
• Transmission is a capacity bottleneck in downlink
• Base station buffers data from Iub
• Future choice – Ethernet/IP transmission
6
Capacity – HSDPA
• HSDPA uses the power margin left over from R99 services
• HSDPA throughput depends on achievable Signal-tointerference and noise ratio (SINR)
• Power allocated for HSDPA effects the throughput largely
Throughput[Mbps] = 0.0039 x SINR^2 + 0.0476 x SINR + 0.1421
25
20
15
SINR dB
PHS  DSCH
SINR  SF16
(1   ) Pown  Pother  Pnoise
10
5
0
0,0
0,5
1,0
1,5
2,0
2,5
3,0
-5
-10
Average cell throughput [Mbps]
7
Load sharing
• Enhances performance by pooling together resources
• Inter Frequency Load Sharing
- Traffic sharing between WCDMA carriers
• Directed Retry to GSM
- Speech traffic diverted from WCDMA to GSM
8
Directed Retry to GSM
Why?
• 3G most beneficial for PS data users
• 3G UEs are becoming more common and
coverage improves constantly
-> free capacity to GSM
• More resources for data users
• Balances load between networks
• No additional investments
9
Directed Retry to GSM - limitations
• Only applicable to speech traffic
• Coverage of UMTS and GSM cells should be same
• Overloading of GSM network possible
• Should not be used if GSM -> UMTS
handovers are in use (ping-pong effect)
• GSM target cell quality not guaranteed (blind ISHO)
-> call drops
• Increased signaling, mobiles not reachable during
LA updates
• Configuration to entire network can be laborious
• 3G users may wonder why they are in GSM
10
Directed Retry to GSM - principles
• Redirection decission based on cell load
(used downlink carrier power)
• After cell load exeeds specified sharing threshold, speech calls
are diverted to GSM network
• Sharing fraction parameter specifies the percentage of directed calls
while the cell load is above the sharing threshold
• Released power can be allocated for PS users
• Sharing parameters can be assigned independently to each cell
Cell load (downlink carrier power)
max carrier power
capacity reserved for HSDPA
this load directed to GSM
sharing threshold
time
Directed Retry active
11
0:
00
1:
00
2:
00
3:
00
4:
00
5:
00
6:
00
7:
00
8:
00
9:
0
10 0
:0
11 0
:0
0
12
:0
13 0
:0
14 0
:0
15 0
:0
0
16
:0
17 0
:0
18 0
:0
19 0
:0
0
20
:0
21 0
:0
22 0
:0
23 0
:0
0
packet
0:
00
1:
00
2:
00
3:
00
4:
00
5:
00
6:
00
7:
00
8:
00
9:
0
10 0
:0
11 0
:0
12 0
:0
13 0
:0
14 0
:0
15 0
:0
16 0
:0
17 0
:0
18 0
:0
19 0
:0
20 0
:0
21 0
:0
22 0
:0
23 0
:0
0
percentage of peak hour traffic
speech
percentage of peak hour traffic
Load sharing - traffic profiles
120 %
100 %
80 %
60 %
40 %
20 %
0%
time
120 %
100 %
80 %
60 %
40 %
20 %
0%
time
12
Load sharing - measurements
• Functionality and different parameters were first
tested in a single cell
• Larger scale test in live network for a three week period
• 20 most loaded cells were chosen for the measurement
• Feature was tested with ”radical” parameters to really find
out how load sharing performed
• A set of key performance indicators (KPI) was defined to
assess the effects of the feature
• DR-success ratio, Speech setup success rate, dropped calls,
admission number, lack of CEs, speech traffic (Erl), PS traffic ….
• Network counters were used to gather information
about the functionality
• Raw data was filtered and manipulated for final results
13
Load sharing - results
• Directed Retry to GSM worked well in overall
• Total of 93117 speech call redirection attempts, 86033
were succesful
• DR-success rate was 92 %, with carefull cell selection
> 95 % success rate possible
Directed Retry to GSM success rate
14
12
Cells
10
8
6
4
2
0
-
5
%
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
% % % % % % % % % % % % % % % % % % %
Success rate
14
Load sharing – results UMTS
KPI
DR - FALSE
DR - TRUE
Difference
Speech setup success rate
99,63
29,29
-71 %
Dropped calls percentage
0,36
0,53
48 %
Speech traffic (Erl)
1,12
0,39
-65 %
PS R99 traffic (Erl)
1,86
2,06
11 %
PS R99 setup success rate
94,41
98,96
5%
PS R99 retainability
94,14
97,97
4%
0,72
0,79
10 %
HS User Thu DL (kbps)
139,24
158,21
14 %
HS User Thu UL (kbps)
33,38
42,58
28 %
HS setup success rate
97,32
97,65
0%
HS completion success rate
79,35
76,36
-4 %
240 706 657
99 813 954
-59 %
153 603
164 147
7%
HS drop %
17,13
20,06
17 %
Admission number
2 949
1 504
-49 %
Failed after admission
3 013
2 789
-7 %
15 733
8 057
-49 %
2 351
817
-65 %
HS traffic (Erl)
CS speech payload (kbits)
HSDPA RAB attempts
NG user down-switches
UL hardware lack
15
Load sharing – results GSM
KPI
DR - FALSE
DR - TRUE
Difference
623 921
820 990
32 %
TCH H_Block %
0,06
0,02
-62 %
TCH T_Block %
0,73
1,66
129 %
TCH RF_Loss %
0,14
0,23
60 %
TCH traffic (Erl)
2,67
3,26
22 %
TCH attempts
16
Conclusions
• Transmission, CEs and HSDPA power allocation
can form a capacity bottleneck
• Load sharing between UMTS and GSM works reliably,
if configured well
• Performance of PS users can be enhanced with
Directed Retry to GSM, at least a little
• 3G traffic still relatively low, it is questionable if
load sharing is needed at this point.
17
Future research
• Load sharing between UMTS carriers
• more sophisticated feature than Directed Retry to GSM
• between UMTS2100 and UMTS900
• two way directions taking into account cell load
• applicable to all services
• Service based handover
18
THANK YOU !
19