Paging Drop Analysis

Paging Drop Analysis

Sadok Ben ALI

Air Interface PCH Load

LAC re-designe

LAC re-designe

Hourly Counters

Daily Counters

LAC re-designe Paging Type1 messsages has been incresed almost to the double due to LAC fusion (More cells per LAC) Number of droped pagging has been incresed proportianly with the paging attempts. Paging Drop

Air Interface PCH Load LAC re-designe

NO Congestion on Air interface

RNC CHR

 After LAC re-designe the PCH load increase up to 8% which is still under the recomded treshold which is 30% This is should be anlyzed on the LAC based

Paging Procedure - Paging Flow in RNC - Bottleneck

The first discarding option in RRC Master is due to processing load reason. Is not the cause based on the counters (No paging drop due ICSU) The second one in MAC-c happens during scheduling . The paging messages stored in buffer can be rescheduled in the next paging occasion (SFN)Paging buffers are 1 each SFN up to the maximum possible number related to DrxCycleLenght. Paging messages are discarded once the buffer time out. Is confirmed By counters M1006C251: PAGING DROP LOW PRIORITY When the L2 entity of RCN drops low priority paging message due to congestion M1006C252: PAGING DROP HIGH PRIORITY When the L2 entity of RNC drops high priority paging message due the congestion

Paging Procedure Buffering - Theory

M = int (window_size div DRXCycleLength) Where window_size is an hidden parameter. This maximum number of those RNC internal attempts also allows to calculate the maximum storage time of paging messages. Using a typical value of window_size = 300 we have: DRXCycleLength=32  M=9; DRXCycleLength=64  M=4; DRXCycleLength=128  M=2; DRXCycleLength=256  M=1.

Drxcycle length Numbe of paging occasion Number of buffer Duration of buffer ms

M

32

127

127

2880

9

64

63

63

2560

4

128

31

31

2560

2

256

15

15

2560

1

Note: Using this coefficient M it is possible to calculate the overall buffering time (e.g. with DRX cycle =320ms a M * DRX cycle = 9 * 320 ms = 2.88 sec of buffering time is possible; the relationship with CN type1 repetition and in type2 RNC must be taken into count.

Conclusion •

The bottleneck at the MAC-c entity, to avoid collision during the scheduling activity a reduction of DrxCycleLength is recomnded:

•  A smaller DrxCycle returns more occasion for UE and more buffering capacity as described in in previous slides.

•

The Congestion is not Hardware is on UMTS paging schedlunig procedure

Annex - Paging Procedure

Paging Procedure Paging Occasion Calculation (1-3) To be camped on the right PI a UE must perform the following procedure: Select the appropriate Secondary-CCPCH Identify the System Frame Number SFN Identify the PI to be read 1. If multiple S-CCPCH are available the UE select one of the S-CCPCH according to the following: S-CCPCH index = IMSI

mod K

where K is the number of S-CCPCH used. If 2 S-CCPCH are used all UEs with even-numbered IMSI will receive their paging on the first S-CCPCH and the UEs with odd-numbered IMSI will receive the paging on the second S-CCPCH. IMSI is given as sequence of digits of type Integer(0..9), IMSI in the formula above is interpreted as a decimal integer number, where the first digit given in the sequence represents the highest order  digit. 2. The equation to identify the exact SFN to monitor is: SFN = (IMSI

div K) mod (DRXCycleLength)

+ n * DRXCycleLength

where n is an integer that varies from “0” to a maximum given by the DRXCycleLength. Note: SFN in literature is known as „Paging Occasion‟

Paging Procedure Paging Occasion Calculation (2-3) The maximum value of the coefficient “n” of the previous formula comes from the DRXCycleLength used as in the following: assuming, for example, a DRXCycleLength=64 frames and considering that SFN range is from “0” to “4095” the maximum possible value of “n” is given by 4095/64=63 This means that the paging occasion never will be greater than 4095 and occurs one per DRXcycle. RNC will use the SFN formula given to individuate the next paging occasion to be transferred to UE. The MAC layer of RNC is responsible to schedule the PI and subsequent message inside the appropriate SFN.

3. Once identified the frame the UE must read the right PI inside the frame. PI = (IMSI

div 8192) mod Np

where Np is the number of PI per frame. In case of collision buffering procedure is performed. The NSN current implementation gives an opportunity of 512 buffers to store the paging messages (one for each paging occasion=SFN up to DRXcycle=5120ms!).

Paging Procedure Paging Occasion Calculation (3-3) When a new paging message arrives and the next paging occasion is already occupied, the paging message is stored in the first free paging occasion belonging to the paging group. The nearest paging buffer place (i.e. the nearest paging occasion) in the paging buffer can then be calculated (neglecting a typical delay of 2 frames to leave time for MCFTOR to schedule) as in the following: Nearest Paging Occasion = (SFN_not_free + k * DRXCycleLength) where k = [1, 2,…M] should get the smallest possible integer value up to the place is found It is possible to calculate for each DRXCycleLength the maximum of coefficient k (M) as in the following: M = int (window_size div DRXCycleLength) Where window_size is an hidden parameter. This maximum number of those RNC internal attempts also allows to calculate the maximum storage time of paging messages. Using a typical value of window_size = 300 we have: DRXCycleLength=32  M=9; DRXCycleLength=64  M=4; DRXCycleLength=128  M=2; DRXCycleLength=256  M=1.

Thank you