DRX x Latency in LTE

4G LTE
, , ,
1

Hi.
When activating DRX in LTE the Latency (OSS KPI) increase but there’s no impact on end user.
Any explanation?

2

There is impact as DL packet needs to wait DRX on to be delivered and if UE is out of sync eNb will request via PDCCH order procedure.
I mean eNB waits for DRX active periods.

3

But with speedtest there’s no impact on Latency.

4

Speed test is initiatied by UE and mormaly with regular DRX setings it should not have impact on speed test perofmances.
It is very important how did you set ul reactivity timer… as longer timer is less impact of DRX to speed test can be expected.

5

Here sre settings:

  • set EUtranCellFDD=L.*,EUtranFreqRelation=1350 cellReselectionPriority 7
  • set EUtranCellFDD=L.*,EUtranFreqRelation=800 cellReselectionPriority 7
  • FeatureState 1
  • set DrxProfile onDurationTimer 7
  • set DrxProfile drxInactivityTimer 14
  • set DrxProfile drxRetransmissionTimer 4
  • set DrxProfile longDrxCycle 9
  • set DrxProfile longDrxCycleOnly 9
  • set DrxProfile shortDrxCycle 9
  • set DrxProfile shortDrxCycletimer 1
6

Hope both the bands are on same spectrum…
If different, we need to set different priorities for different spectrum bands.
Eg : High priority for Large Spectrum and second priority to other spectrum
Ex : if FDD 20 - 7 priority and FDD 10 - 6 Priority

8

What about timers settings?

9

Hi LTE Experts.
If there is high DL latency then apart from checking DRX settings, packet loss at router end by pinging IP any other check can be done?

10

Have you narrowed down which link or adding is adding most latency? Think in Huawei you can check latency directly b/w eNB and UE by pinging UE IP.

Anyway here are some ideas:

1- without core guys support (simple but less accurate tests):
1a- test with different servers- sometimes latency is added by public servers used for testing
1b- test different times of the day and each time start test with UE in airplane mode for good 15 20min - this will help in confirming if latency is due to scheduling at eNB. Airplane mode helps reset UE context across all nodes.

2- with core guys support (more complicated but accurate):
Put traces in UE, eNB, one of the nodes in transport network (decide which is convenient), PGW and a server (can be a FTP server sitting behind PGW just like internet or Application server).
Start a ping from either UE or FTP server.
Track packets at each node
Check the timing of packets and see which interface has added most latency.
You need to make sure that timestamps on all these nodes are same (I.e. using same timing clock)
In some cases, a specific UE pool can have different routing mechanism adding latency to packets. This test can show you that

11

Check by traceroute to find which route is contributing to major latency.

12

Good suggestion.
But please note Tracert is IP layer.
Almost entire telecom infrastructure is transparent to traceroute between UE and server.

13

Thanks all!
So latency majorly need core side debugging.

Any parameter fine tuning which can help?

14

Latency in lte can be tried by pdcp discard timer.
In non gbr normal they set infinite.
In VoLTEe around 150 to 300 ms.

15

Are you suspecting scheduling issues as well?

16

Intermittent kind of ping loss.
With high latency.

17

If it is intermittent, there could be some threshold issue or some link issues as well.

18

First please check whether backhaul is error free.
If backhaul is ok, then radio link parameters may be checked.

1 Like
19

Does DRX can be issue here?
I mean DRX switch OFF can help?

20

You can disable and observe.

21

With Connected Mode DRX, the UE saves battery by listening occasionally to the PDCCH. This increases the DL one-way latency if the UE is into sleep mode as it takes longer to transfer a packet to the UE.

1 Like