You are on page 1of 5

RRC Connection Setup Rejection due to Radio Link Failure

1) Trouble description:

Filed engineer noticed that for some NodeBs, the RRC KPI decrease from time to time,
more than it should normally do.

So customer asked why there are so big variations in RRC KPI.

2) Trouble Analysis

After the analyzing data from the RNC, the cause for the degradation is
VS.RRC.Rej.RL.Fail.

From the CHR data, we found most of RL fail cause is unspecified:


3) Trouble description:

Filed engineer noticed that for some NodeBs, the RRC KPI decrease from time to time,
more than it should normally do.

So customer asked why there are so big variations in RRC KPI.

4) Trouble Analysis

After the analyzing data from the RNC, the cause for the degradation is
VS.RRC.Rej.RL.Fail.

From the CHR data, we found most of RL fail cause is unspecified:


The RADIO LINK SETUP FAIL was sent by NodeB, so we need to investigate NodeB
side.

After analyzing the NodeB mainboard log, it was found that the above problem was
caused by Dynamic CE allocation.
Dynamic CE is a software that schedule DSP resources periodically to drive users to
achieve a total high and balanced throughput.
There are 2 UL DSPs on BBU, DSP1 and DSP4.
DSP1 controls 4 ASIC chips, and DSP4 controls 2.
When users built on DSP4 cannot grab NDC resources from DSP1, and within
DSP1,users built on one chip cannot grab NDC resources from another.
This is why RL failures happen to users that setup on DSP4.
Another module, RM (Resource Manager), guarantees each user with minimun
resources and Dynamic CE schedule each accessed user to make most of NodeB
resources.
When a new user tries to access,Dynamic CE module may need to release some
resources from other users.
When a large amount of users try to access and exhaust NDC resources, and action
usch as Add or Reconfigure that need to grab NDC resources between ASIC chips
may fail due to Dynamic CE way of fucntion.

Conclusion is that RL failures occur because of the way Dynamic CE function


algorithm.

Solution:
- The solution was to add an EBBC board on the affected NodeBs, to increase
capacity.
- Another solution was to disable Dynamic CE function, but this increased the risk
of UL congestion.

You might also like