PDA

View Full Version : The problem of traffic to the RNC E******* 3810 and RAB Setup PS



GrayDog
2012-01-29, 03:50 AM
Hi all! If someone has a similar experience, please help.
In the busy hour of HS users at the RNC (around 19:00-21:00) there is a significant reduction in packet traffic. At the same time increasing the percentage of unsuccessful RAB Setup PS. Counters for the reasons - empty (Lack UL / DL Hw, etc.). I think the problem lies in the PS Core (SGSN may or GGSN), but still can not find.
Grateful for any ideas.

roloramos
2012-01-31, 05:11 AM
Hi Friend..
What about the MP, DC, CC, PDR load??

s52d
2012-01-31, 08:37 AM
Hi!

How about licencing capacity?
A while ago (P7 times?) there were serios problems when licence limits were approached.

BR
s52d

GrayDog
2012-01-31, 03:24 PM
Hi Friend..
What about the MP, DC, CC, PDR load??
Hi,
MainProcessor load High, but counter pmNoRejRrcConnMpLoadC is empty. I understand that in case of overload MP - will be failures on RRC pmNoRejRrcConnMpLoadC?
DC, CC, PDR load - How check problem (any counter or formula load) ? Now i read WCDMA RAN P7 Troubleshooting - about DC, CC, PDR.

Traffic RNC - with an increase in the number of HS users, throughput IuB reduced to RNC
24273

GrayDog
2012-01-31, 04:34 PM
Hi!
How about licencing capacity?
A while ago (P7 times?) there were serios problems when licence limits were approached.


Hi!
RNC SW - P7, License about 925 Mbitps, RNC Iub Thr ~ 800 Mbitps
From experience, licensing restrictions cut traffic ( limiting the RNC Iub traffic ), but does not lead to the concavity of curve

aygo
2012-01-31, 07:45 PM
Hello, Sergey

how to RNC connected to NodeB? and which model traffic as HSDPA users?

roloramos
2012-01-31, 09:42 PM
The table on Eniq DB is:
DC_E_RAN_DCDEVICE_RAW A

The KPI are:
SUM(pmSamplesMeasuredLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredLoad)/SUM(pmSamplesMeasuredLoad) END AS MP_LOAD

SUM(pmSamplesMeasuredCcSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredCcSpLoad)/SUM(pmSamplesMeasuredCcSpLoad) END AS CC_LOAD

SUM(pmSamplesMeasuredDcSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredDcSpLoad)/SUM(pmSamplesMeasuredDcSpLoad) END AS DC_LOAD

SUM(pmSamplesMeasuredPdrSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredPdrSpLoad)/SUM(pmSamplesMeasuredPdrSpLoad) END AS PDR_LOAD

The QUERY is:

SELECT
M.RNC,
M.DATETIME_ID,
MP_LOAD,
CC_LOAD,
DC_LOAD,
PDR_LOAD
FROM
(
select
A.RNC,
A.DATETIME_ID,
CASE WHEN SUM(pmSamplesMeasuredLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredLoad)/SUM(pmSamplesMeasuredLoad) END AS MP_LOAD
from DC_E_RAN_LOADCONTROL_RAW A
WHERE DATETIME_ID>='2011-11-10 01:00' AND DATETIME_ID<='2011-11-22 23:00' and RNC='rnc_name'
GROUP BY DATETIME_ID, RNC)M, (select
A.RNC,
A.DATETIME_ID,

CASE WHEN SUM(pmSamplesMeasuredCcSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredCcSpLoad)/SUM(pmSamplesMeasuredCcSpLoad) END AS CC_LOAD
from DC_E_RAN_CCDEVICE_RAW A
WHERE DATETIME_ID>='2011-11-10 01:00' AND DATETIME_ID<='2011-11-22 23:00' and RNC='rnc_name'
GROUP BY DATETIME_ID, RNC)C, (select
A.RNC,
A.DATETIME_ID,
CASE WHEN SUM(pmSamplesMeasuredDcSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredDcSpLoad)/SUM(pmSamplesMeasuredDcSpLoad) END AS DC_LOAD
from DC_E_RAN_DCDEVICE_RAW A
WHERE DATETIME_ID>='2011-11-10 01:00' AND DATETIME_ID<='2011-11-22 23:00' and RNC='rnc_name'
GROUP BY DATETIME_ID, RNC)D, (select
A.RNC,
A.DATETIME_ID,
CASE WHEN SUM(pmSamplesMeasuredPdrSpLoad)=0 THEN 0 ELSE
SUM(pmSumMeasuredPdrSpLoad)/SUM(pmSamplesMeasuredPdrSpLoad) END AS PDR_LOAD
from DC_E_RAN_PDRDEVICE_RAW A
WHERE DATETIME_ID>='2011-11-10 01:00' AND DATETIME_ID<='2011-11-22 23:00' and RNC='rnc_name'
GROUP BY DATETIME_ID, RNC)P
WHERE M.RNC=C.RNC AND M.DATETIME_ID=C.DATETIME_ID
AND M.RNC=D.RNC AND M.DATETIME_ID=D.DATETIME_ID
AND M.RNC=P.RNC AND M.DATETIME_ID=P.DATETIME_ID

GrayDog
2012-02-01, 03:14 AM
Thanks roloramos !
That's what I got
24295

roloramos
2012-02-01, 03:16 AM
thresholds used and recomended for E*******:

- mMP Load……………………………………Less than 75%
- RANAP/RNSAP MP Load………………….Less than65%
- SP (Dc Load)……………………………….. Less than85%
- SP Cc, Pdr) Load……………………………Less than60%
- RNC Active Users……………………………Less than90% of licensed capacity
- RNC Iub Throughput……………………….. Lessthan 90% of licensed capacity, but its value depends of the other thresholds

GrayDog
2012-02-01, 03:43 AM
Hello, Sergey
how to RNC connected to NodeB? and which model traffic as HSDPA users?

Hello, RNC connected to NodeB over ATM,
about traffic model is difficult for me to answer, can you describe the necessary parameters of the model or how to define it?

s52d
2012-02-01, 04:58 AM
Hi!
RNC SW - P7, License about 925 Mbitps, RNC Iub Thr ~ 800 Mbitps
From experience, licensing restrictions cut traffic ( limiting the RNC Iub traffic ), but does not lead to the concavity of curve

Hi!
I recall simmilar pictures. When licence triggered, traffic was halved and oscilating. Ugly and nasty.
We upgraded HW and SW, so we are far from limits.

BR
s52d

GrayDog
2012-02-03, 04:08 AM
thresholds used and recomended for E*******:

- mMP Load……………………………………Less than 75%
- RANAP/RNSAP MP Load………………….Less than65%
- SP (Dc Load)……………………………….. Less than85%
- SP Cc, Pdr) Load……………………………Less than60%
- RNC Active Users……………………………Less than90% of licensed capacity
- RNC Iub Throughput……………………….. Lessthan 90% of licensed capacity, but its value depends of the other thresholds


true, RNC in a high load (overload), but in the following graph shows that at higher load can have a normal form of traffic (PDR_LOAD) and do not have any problems with failures RAB_Setup_PS. These changes are related to configuration changes in PS Core, but it is not clear what leads to such results. So, I guess, and what to look for the problem in PS Core (or IP BackBone).
24354

GrayDog
2012-05-31, 02:05 PM
The problem resolved after reducing the number of NodeB to the RNC and fix problems on the GGSN. Problems in the GGSN creating additional load on the RNC. It is important that the 3810 license capacity values ​​IuB 800Mbps or 900Mbps is almost impossible to use because of a specific traffic model, which does not match with the model by default. In fact, a RNC limit of about 600 - 700 Mbps.

shamim23
2012-06-01, 12:58 AM
Thanks Gray. Can you please tell us more about the problem faced in GGSN?