1. 1 TWAREN 混合式網路 監控系統之設計與實作 國家高速網路與計算中心 網路服務組

2. 2 Outline  Introduction  Motivation  Issues  Design  Implementation  Future works

3. 3 INTRODUCTION

4. 4 About TWAREN  TWAREN (TaiWan Advanced Research & Education Network) network construction was completed at the end of 2003 and started its operation and service in the beginning of 2004.  In its initial phase, IP routing was the main service provided.  The network management programs coming along with the purchase of network equipments, including CIC, Webtop, CW2K, HP Openview, HP NNM and other solutions.

5. 5 Initial phase of TWAREN Taipei Taichung Tainan Hsinchu ASCC NDHU NCTU NTHU NCHUNYSU NCU CCU NCU NCCU TWAREN 10GE STM-64/OC-192 STM-16/OC- 48 GE C7609 C6509 GSR NTU C6509 NHLTC C6509 NTTU C6509 EBT10GE MOECC C6509

6. 6 Initial phase of NMS 37502600 15454 15600 25227609 NAM CW2K (DFM) NNMCTM Cisco Info Center Remedy Help Desk ISM Notification 12416 Trap PING Polling Trap PING Polling PING Polling Trap CLI API Gateway Probe HTTP FTP SMTP DNS WebTop

7. 7 Phase 2 of TWAREN  TWAREN was adapted for more protection methods and better availability at the end of 2006, called TWAREN phase 2.  Tens of optical switches and hundreds of lightpaths were then served as the foundation of the layer 2 VLAN services and the layer 3 IP routing services.  In 2008, tens of VPLS switches were further incorporated to provide additional Multi-point VPLS VPN service.  The layer 1 lightpaths can be protected by SNCP, layer 2 VLAN by spanning tree recalculation and layer 2 VPLS by fast reroute technology.  All these improvements transform TWAREN phase 2 into a true hybrid network capable of providing multiple layers of services and high availability.

8. 8 Architecture of TWAREN phase 2 STM64 STM16 10GE GE 6509 7609 15454 NTU 6509 7609 15454 NCU 6509 7609 15454 NSYSU 6509 7609 15454 NCHU 6509 7609 15454 NCTU 6509 7609 15454 NTHU 7609 15454 ASCC 6509 NCCU 6509 7609 15454 NCKU 6509 7609 15454 CCU 7609C 15454 Taipei 15600 12816 NCHC 7609C 15454 15600 12816 MOEcc Hsinchu 7609C 15454 15600 12816 NCHC Tainan 7609C 15454 12816 NCHC Taichung 7609 NCNU 7609 15454 NIU 6509 7609 15454 NDHU 3750 6509 3750 NHLTC 6509 3750 NTTU

9. 9 MOTIVATION

10. 10 Why need new NMS?  The architecture of TWAREN phase 2 became more and more complicated.  Since TWAREN phase 2 has more protection methods, a single point of hardware or circuit failure will not interrupt the service level provided to the end users.  The initial phase of NMS was no longer competent for the hybrid network anymore because it is hard to determine and predict the correlation between failures and affected services.

11. 11 Requirements for new NMS  Automatically determine the correlation between failures, affected services, affected customs and severity level on this highly safeguard network.  Provide single integrated visual user interface.  Use integrated database, logs, message flows and exchange protocols.  After several surveys, we decided to develop a new NMS which be suitable for monitoring all services provided by TWAREN phase 2.

12. 12 ISSUES

13. 13 Uncertainty of SNMP implementation  There are some different implementations of the SNMP TRAP/MIB among equipments of same brand.  The SNMP OIDs or the return values may vary between OS upgrade on the same equipment and are usually hard to reveal beforehand.  Therefore, the system must be designed in a way such that these changes can be accommodated with minimal modifications.

14. 14 The lack of skillful programmers  Our programmers are the same guys with the members of operating team.  We are not professional programmers and have not accordant programming language.  The system must be partially available and operational during the early phase of its development such that it can evolve along with the real needs.  So, an unified standard of communication between different modules is necessary

15. 15 Huge historical data and computing  For minimizing the false positive and false negative rate, baseline thresholds would have much better quality when they are dynamically generated from historical data.  Therefore, we need to store sufficiently large historical data sets and to have very high efficiency to retrieve the data back while calculating those thresholds.

16. 16 Automatically determine affected services and customs  TWAREN phase 2 inherently has the ability to guard against a single point of hardware or circuit failure, so the failure is less likely to affect the actual service provisioning.  An intelligent management system which is able to determine the scope of failure affected service will reduce the management cost.

17. 17 DESIGN

18. 18 1 st Stage System Architecture Current Status DB Long Term DB Monitor Objs Data Collectors Traps MIBs Syslogs Net flows Telnet/SSH Fault Detection Threshold DB Case/Action DB GUI & Ticket System Threshold Analyzer Fault Location Auto Action Control API Report System TL1 Mirror Interactive Passive

19. 19 Relationship of Data Tables Component People Location Unit Vendor …., etc Basic Data Tables Circuit VLAN Services VPLS Services ONS Light Path ONS Cross Connection …., etc Relationship Tables

20. 20 IMPLEMENTATION

21. 21 Current monitor objects  Trap monitor  Used interfaces, BGP, etc.  Environment of equipment room  Temperature (auto threshold), Voltage  Statuses of equipments  Temperature, CPU, RAM, FANs, Power-Supply  BGP peering with other networks  Statuses, Number of exchanged routes (auto threshold), Utilization analysis  Performance monitor  End to End RTT (auto threshold), End to End Packet Lost Rate (auto threshold), End to End Availability  Throughput  Backbone (auto threshold), Designate interfaces  Top N  Bytes, Flows, Packets  Routes monitor  The routes of customs (exact comparison)  VPLS VPN  Throughput of CE side, MACs of VPN, Status of MPLS PW  Optical Network  Current topology of lightpaths  VLAN  Current topology of VLAN

22. 22 Future works  Combine all developed monitor objects with single integrated visual user interface.  Enhance the monitoring of optical, VPLS and VLAN networks.  Automatically determine the fault location, root cause and affected scope.  Minimize the false positive and false negative rate.