PS drop due to User Inactivity

[Spider]

Hi Experts,

During an audit on E// network we observed a lot of PS drops on 3G layer due to User inactivity, from traces we have a lot of IU_Rel_CMD triggered from SGSN and RNC sometimes, i have tried to correlate with some timers but no success, anyone can explain this behaviour and suggest some recommendations to reduce this kind of drops?

Thanks :)

[gprastomo]

Hi, this is due to 0 activity during specific periods.
To reduce this, you can activate FACH PCH, since changing states will restart the timers, or you can prolong the timers but it will consume resource which may appears congestion.

[Spider]

Thanks gprastomo, any suggestions for the optimal settings for the timers? i think this parameter change is a big issue since it's a tradeoff between system capacity and end user perception.

[kentck86]

I was experiencing this in huawei network too. Once the UE hitting the threshold data rate for Event 4B (Lets say 64bps), then it will trigger to inactivity timer of 20s(For huawei default setting). Within the 20s if the data rate wasnt hit the threshold of event 4a (lets say 1024bps), the radio link will be released with the cause of Userinactivity. This feature kickoff the user which doesnt perform any data activity, by this way it could giving more spare of resources for new connection. There were pros and cons for the lengthening of the timer or state transition.;)

[gprastomo]

I recommend to activate fach and pch, which can drastically improve this inact. But remember set the H2F to 2-5s, F2P to 2-5s, in the future when you have huge number of smartphone, you can prolong this inact timers.
Also dont forget to set the activity factor for IuPS reservation to 40%

[Spider]

Actually in our network we have a lot of Blackberrys which are using a lot BBM service !!! Indonesian culture, anyway, Actually the inactivity Timer is set to 10s. i checked and The state of FACH and PCH is enabled for all cells. Really confused about this issue, i think even if we increase inactivity Timer in the RNC it will not be taken into consideration since the fast dormancy regulates the procedure. Any comments?

[regfaycat]

I don't think the connection release due to user inactivity should be classified as a drop, since it provides a way to use the network resource effectively. Activation of CELL_FACH and CELL/URA_PCH modes can indeed help reducing the number of state transitions between connected and idle mode, nevertheless, the majority of smart phones implement a proprietary fast dormancy mechanism which will initialize the release of IU regardless of the standardized state transitions procedures. The inactivity timer setting changes won't have any influence on the UEs in question.

[gprastomo]

The user inactivity release shouldnt be as a drop, but if the value is too high, the user perception will think that the connection is always drop, so we need to manage.
If you set the inact timer to only 10 second, it s too fast so for for smartphone user, which release the call due to this timer will automatically reconnect, and this will make huge of unnecessary signalling. I believe with 10 s of inactivity timer , you ll get overload soon even the traffic is small.
Fast dormancy is different from the hspa rrc state. Fast dormancy is mobile managed its inactivity by sending the SCRI, (usually the timer is different for each mobile vendor).
So you need to see :
1. for fast dormancy enabled UE, you need to manage the PCH to Idle inactivity timer, usually its better to take > 300 s
2. For non fast dormancy ue, we need to prolong the inactivity timers to > 300 s, to reduce the inactivity.

[Spider]

Thanks for the reply, first step i'll try to change the inactivity timer and see the impact. I'll keep you informed by the results, hope also that the MP load will decrease after this operation.

[gprastomo]

Yes, just monitor the call attempt, or RRC att, then also monitor the CPU Load or processor which handles the RRC

[Spider]

Back again, actually we changed the Inactivity timer on RNC level from 10s to 30s and the RAB PS drop increased by 0.8% to reach 1.8% in average, from probing we see less User inactivity , the MP load was quite stable after the change, the number of SR for upswitch from FACH to HS decreased, any comments regarding this behavior?

The formula for our PS drop is the following: RAB Drop Ps =100*((pmNoSystemRabReleasePacket)/(pmNoRabEstablishSuccessPacketInteractiveHs+pmUpswitchFachHsSuccess+(pmNoNormalRabReleasePacket-pmNoNormalRbReleaseHs)+(pmNoSystemRabReleasePacket-pmNoSystemRbReleaseHs)))