Professional Documents
Culture Documents
LEGAL INFORMATION
By accepting this certain document of ZTE CORPORATION you agree to the following terms. If you do not agree to the following terms, please notice that you are not allowed to use this document. Copyright 2014 ZTE CORPORATION. Any rights not expressly granted herein are reserved. This document contains proprietary information of ZTE CORPORATION. Any reproduction, transfer, distribution, use or disclosure of this document or any portion of this document, in any form by any means, without the prior written consent of ZTE CORPORATION is prohibited. and are registered trademarks of ZTE CORPORATION. ZTEs company name, logo and product names referenced herein are either trademarks or registered trademarks of ZTE CORPORATION. Other product and company names mentioned herein may be trademarks or trade names of their respective owners. Without the prior written consent of ZTE CORPORATION or the third party owner thereof, anyones access to this document should not be construed a s granting, by implication, estopped or otherwise, any license or right to use any marks appearing in the document. The design of this product complies with requirements of environmental protection and personal security. This product shall be stored, used or discarded in accordance with product manual, relevant contract or laws and regulations in relevant country (countries). This document is provided as is and as available. Information contained in this document is subject to continuous update without further notice due to improvement and update of ZTE CORPORATIONs products and technologies.
ZTE CORPORATION
Address: Address: NO. 55 Hi-tech Road South ShenZhen P.R.China 518057
http://dms.zte.com.cn TSM.Aftersales@zte.com.cn
Website: Email:
Revision History
Product Version Document Version R1.0 Serial Number Reason for Revision First published
Author
Date 2012-10-10 Document Version R1.0 Prepared by Chen Yanhao, Cai Wei Reviewed by Chen Taiming Approved by Chen Taiming
II
III
TABLE OF CONTENTS
1 2 2.1 2.2 2.3 2.3.1 2.3.2 2.3.3 2.3.4 3 3.1 3.2 3.3 3.3.1 3.3.2 3.3.3 4 Background..................................................................................................... 1 RCB CPU Load Monitoring Solution & High Load Handling Solution ........ 2 Principle & Calculation of RCB CPU Load Increase .......................................... 2 RCB CPU Load Monitoring Solution ................................................................. 3 Solution to High Load of RCB CPU ................................................................... 4 Adjusting Number of Sites Among RCP Modules ............................................. 4 RCB Board Expansion ...................................................................................... 4 Inter-RNC Sites Migration ................................................................................. 4 Others............................................................................................................... 4 RUB CPU Load Monitoring Solution & High Load Handling Solution ........ 6 Basic Principles of RUBCPU Load Control Policy ............................................. 6 RUB CPU Load Monitoring Solution ................................................................. 6 Solution to High Load of RUB CPU ................................................................... 6 Balancing CPU Load Among RUBs of the Same Resource Shelf ..................... 6 Balancing CPU Load Among RUBs of Different Resource Shelves .................. 7 Performing RUB & Resource Shelf Capacity Expansion ................................... 7 Summary ......................................................................................................... 9
IV
TABLES
Table 2-1 Control-Plane Service Model ................................................................................ 2 Table 2-2 RCB CPU Load Monitoring KPIs & Threshold ...................................................... 3 Table 3-1 Counters of Discarded Messages......................................................................... 6
Background
RCB, i.e., RNC control-plane processing board, consists of two CPU modules. In V3.11, RCB can fulfill different functions (i.e., CMP and DMP) by running different versions. CMP processes the common messages of the control plane, while DMP processes the dedicated messages of the control plane. RUB consists of one CPU and 15 DSPs. CPU exchanges messages with other boards and manages the DSPs. With the increase in users and service applications, the signaling traffic and RCB load are increased. Besides, with the increase in signaling traffic, the messages exchanged among the CPU of RUB, other boards, and the DSP of the current board increase continuously, so does the CPU load of the RUB. When the load reaches the threshold, network congestion will occur and network KPIs will become worse. This will affect end users service experience There was the situation that call setup failure occurs due to high CPU load of RCB and RUB. If the CPU load of RCB is too high, there will be system error, which will lead to more serious problems.
RCB CPU Load Monitoring Solution & High Load Handling Solution
Principle & Calculation of RCB CPU Load Increase
The increase in RCP CPU load is directly related to BHCA. The relation between RCP CPU load increase and BHCA increase is similar to linear relation. Therefore, whether the current CPU load increase is reasonable can be ascertained by comparing BHCA load curve with RCP CPU load curve. CPU utilization ratio = Expected equivalent BHCA / BHCA supported by RCP
2.1
Table 2-1
Control-Plane Service Model Value (Example) Unit Numb er of times/ busy hour Weight Counter No. (U9.3) C310090252+C310090274 +C310090275+C31009027 6+C310090277+C3100902 80+C310090293+C310090 309 C310322216+C310322217 +C310322218+C31032223 2+C310322233+C3103222 34+C310332558+C310332 514+C310332536+C31033 2547+C310332503+C3103 32525+C310342761+C310 342780+C310342807+C31 0342826+C310342837+C3 10342838+C310342839+C 310342840+C310342841+ C310342842+C310342843 +C310342849+C31034285 0+C310342851+C3103428 52+C310342853+C310342 854+C310342855+C31034 2861+C310342862+C3103 42863+C310342864+C310 342865+C310342866+C31 0342867+C310342873+C3 10342874+C310342875+C 310342876+C310342877+ C310342878+C310342879 +C310352945+C31035295 2+C310352959+C3103528 85+C310352886+C310352
Control-Plane Service Model Number of RAB assignments for busy-hour CS/PC service per user
100%
20
35%
Value (Example)
Unit
Weight
Counter No. (U9.3) 887+C310352888+C31035 2889+C310352890+C3103 52891+C310352897+C310 352898+C310352899+C31 0352900+C310352901+C3 10352902+C310352903+C 310575251+C310575252+ C310575255
0.17
35%
C310404201+C310404203
31%
C310414096+C310414097 +C310414098+C31041409 9+C310414100+C3104141 01+C310414102+C310414 103+C310414104+C31041 4105+C310414106+C3104 14107+C310414130+C310 414131+C310414132+C31 0414133+C310414134+C3 10414135+C310414136+C 310414137+C310414138+ C310414139+C310414140 +C310414141+C31041690 1+C310416902+C3104169 03+C310416904+C310416 909+C310416910+C31041 6911+C310416912+C3105 75256+C310575257 C310080005+C310080006 +C310080007+C31008001 2+C310080013+C3100800 14+C310080015+C310080 016+C310080017+C31008 0018+C310080019+C3100 80020
Number of times of busy-hour RRC per user without RAB assignment Number of times for processing ALCAP
44.7%
15
15%
C310930000
In Table 2-1, the figures in red are examples of services occurring in busy hours. They can be modified according to the operators requirements or the statistics in the existing network. Equivalent BHCA requirements = 31 + 200.35 + 0.170.35 + 40.31 + 30.447 +150.15 = 14.89. In the above formula, the figures in red are corresponding to those in Table 2-1, and they can be modified according to the actual situation.
CPU utilization ratio = Equivalent BHCA / BHCA supported by RCP (14.89/250k for the above example) Note: The calculation in V3.11 is changed and will be updated later.
2.2
Table 2-2
Monitoring KPIs
Average CPU load Peak CPU load C311340001
Granularity
15 minutes 15 minutes
Alarm Threshold
50% 70%
C311340000
When the average CPU load of RCB reaches 60%, it is necessary to perform load balance or capacity expansion to the RCB board. When it reaches 70%, which is a really dangerous, this situation needs to be handled immediately.
2.3
2.3.1
3.
4.
Observe the busy-hour KPIs and RCP CPU load after the modification and see whether the load among RCPs is balanced.
Note: After the home module of the sites is modified to the RCP module, it is necessary to observe whether this modification causes imbalance to other resources, e.g., number of selective preference shelf subsystems.
2.3.2
2.
Note: After the RCB board is added, the number of sites should be balanced among selective preference shelf subsystems, so as to avoid service inconsistency among resource shelves.
2.3.3
2.
3.
4.
Note: The neighbor cells need also to be adjusted in the migration process. After the migration, the number of sites should be balanced among the selective preference shelf subsystems in the target RNC.
2.3.4
Others
1. If the current RNC version is V3.09, it can be upgraded to V3.11 to reduce RCB load. In the case of emergency, such as sudden high-load and grand festivals, master/slave RCB can be configured as non-master/slave mode, so that the number of available RCP modules can be increased to reduce the load. For any problems about the above operations, please contact the chief engineers in the headquarters.
2.
3.
RUB CPU Load Monitoring Solution & High Load Handling Solution
Basic Principles of RUBCPU Load Control Policy
There are two thresholds on RNC, i.e., RupCongestThr and RupResumeThr (RupResumeThr < RupCongestThr). When RUP detects that the average CPU seizure rate is greater than RupCongestThr for RupJudgeTime seconds, this RUP will become CPU load control state. After an RUP becomes load control state, the RUP will receive the CCCH signaling (mainly RRC connection requests) of the RUP by RupAcceptRate. If the load of the current RUP within RupJudgeTime seconds is still not reduced to RupResumeThr, this RUP will access service by RupAcceptRateRupAcceptRate (powers of N), and so on. If the average load is reduced to RupResumeThr, this RUP becomes normal operation state. The default value of RupCongestTh is 95%. When RupCongestThr is configured as 100%, the RNC will not start this load control function. After the load control policy is started, the discarded CCCH signaling is mainly RRC signaling. In this case, users cannot access the network. The discarded messages can be observed through the following RNC counters.
3.1
Table 3-1
Counters of Discarded Messages Measurement Type Cell transmission quality statistics Cell transmission quality statistics Measurement Subtype CCCH signaling transmission quality statistics CCCH signaling transmission quality statistics Counter No. C31053599 4 Counter Name Number of uplink CCCH signaling discarded Number of uplink CCCH signaling
C31053599 5
3.2
Monitor the C311340001 counter (%) and collect statistics by the granularity of 15 minutes. Monitor average RUB CPU load when it reaches 60%. When the average RUB CPU load reaches 80%, perform load balance or capacity expansion to the RUB board.
3.3
3.3.1
MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock MasterUnBlock
11 12 13 14 15
30 9 29 0 21
23 17 27 37 45
Status: If the DSP status is abnormal, check whether the DSP is blocked and whether DSP service IP is configured. CellNum: If the number of cells is imbalanced, block/unblock all cells under the preference shelf subsystem in batch. If the number of services is low on some DSPs, contact the chief engineers in the headquarters. After the above operations, collect the statistics of ShowHpmuInfo and observe whether the DSP status is normal, whether the number of cells is balanced, and whether the CPU load of the RUBs in the shelf is balanced.
3.3.2
If the inter-shelf RUB CPU load is imbalanced, modify the selective preference shelf subsystem of the site from high-load subsystem to low-load subsystem. Block/unblock the cells of the modified site after the modification. Observe whether the CPU load of the RUBs among different shelves has been balanced,
3.3.3
10
Summary
CPU load monitoring of RNC is a long process. With the increase of service volume, it needs to be done continuously. It is necessary to keep monitoring KPIs and analyzing system load to locate and adjust the RNC close to the overload threshold, so as to avoid network KPI decrease, the influence upon users experience, and system error.
11