Wednesday, 4 November 2015

S1 HO Procedure Duration Analysis

1 Description

In a project customer analyze LTE network and find that the S1HO procedure duration is much longer than reference value in other operator.

2 Latency Analysis

 

2.1 S1HO latency definition

Based on operator xxx, S1HO latency is defined as the time difference between Measurement Report Message and RRCConnection Reconfiguration Complete. It is the time from step1 to step14 in this figure, and the common value is less than 300 ms.
 

2.2 Why is the S1HO latency from customer is longer than 5s?

Here is a S1HO procedure trace. From the signal time stamp, we can see that from MME receive Modify Bearer Response message to S1AP-UE ContextReleaseCommand message is around 5s. This latency is the main latency for S1 procedure latency.

 
 

2.3 Does the latency can be changed?

In fact, protocol TS 23.401 defines a resources management timer for MME to manage the resources in source eNodeB. The timer will start after the MME received Handover notify form target eNodeB and will try to release the source in source eNodeB after the timer expire.
This timer can be changed in MME. It can be changed to a minimum 100ms and a maximum 150s value. The currently value for this timer is 5s. That is the reason for the S1HO procedure long duration.
 

  2.4 Does the timer affect service?

For the S1HO procedure, it includes 3 phases: HO preparation, HO execution and HO complete. Here is the full procedure for S1HO:
 
In fact, after HO execution, UE has synchronized to the target eNodeB, UE service can continue in the target eNodeB after it active bearer in target eNodeB. HO complete phase will clear the original resources occupation in source eNodeB.
For the resource release timer, if it is set shorter, the advantage is: system can release the resources in source eNodeB early for new UE access, it can improve resource utilization; however the disadvantage is: S1HO take a risk that if UE fail to active bearer in target eNodeB after HO execution, it will not be able go back to the source eNodeB and can result in call drop; if it is set longer, the advantage is: it make UE service more stable but the disadvantage is reducing resource utilization.

3 Conclusion

1.    Compare with operator xxx, S1HO latency definition from BASE is different. Based on the definition of operator xxx, the typical value is less than 300ms, and currently the S1HO latency for BASE is with this range.
2.    Based on BASE definitioncurrently the S1 HO long latency is caused by the resource release timer (currently setting is 5s) in MME.
3.    The resource release timer in MME can be changed from 100ms to 150s.
4.    For the resource release timer, if it is set shorter, the advantage is: system can release the resources in source eNodeB early for new UE access, it can improve resource utilization; however the disadvantage is: S1HO take a risk that if UE fail to active bearer in target eNodeB after HO execution, it will not be able go back to the source eNodeB and can result in call drop;
The resource release timer setting need compromise the resource utilization and service stability, so recommend 5s for currently LTE network.

Budi Prasetyo

About Budi Prasetyo

All About LTE

Subscribe to this Blog via Email :