This is an overview of the concept of Telecom Performance Management System for Tier-1 and Tier-2 operators.
The main features of this concept are:
- full coverage of all TMN model layers
- the ability for extension to full FCAPS model
- integration with any third-party systems
- distributed and fault-tolerant database model optimized for very large data volumes
2. Operators need PM system to: Predict, analyze and investigate network and service performance degradations Generate and present network and service performance reports to company management Forecast network and service performance in case of events (Exhibitions/Trade Shows, New Year, Olympic games) or new product launches Control compliance with SLA on outsourced equipment October 2010 2 TPMS: System Description
3. General requirements for PM system - 1 Near real-time system Support different data sources like performance counters, CDRs, probes, field/drive test results Scalable for any volumes of input data and retention periods System availability 99,999% Flexible for customization and extension Have open southbound and northbound interfaces Support object-level and domain-level security October 2010 3 TPMS: System Description
4. General requirementsfor PM system - 2 Support multi-vendor, vendor-dependent, multi-service and service-dependent models for data and hierarchy. Support a service-network relation Keep history of changes of network hierarchy, KPIs and reports Support standard telecom functions and methods like Busy Hour, DAV, Erlang etc. Flexible for extension with user-defined functions. Support data forecasting and profiling October 2010 4 TPMS: System Description
5. High-levelSystem architecture As most other systems PM system contains: RAW data collection and parsing layer Data storage and managementlayer Application layer Presentation layer (User interface) October 2010 5 TPMS: System Description
6. RAW data collection and parsing Collect data using FTP, SNMP, CORBA, X.25, SQL, custom scripts Store collected data in input files Unpack files (if needed) Rename files to unified file name (if needed) Identify corrupted files Feed files to parsers Store processed files (may be needed for future data re-load) October 2010 6 TPMS: System Description
7. RAW data collection and parsing Dump files to unified format Process variable file structure and contents Un-peg data Validate and filter data (formula-based) Normalize data Aggregate, accumulate and enrich data Collect and report it’s own performance counters October 2010 7 TPMS: System Description
8. Data storage and managementlayer Data warehouse based on industrial standard DBMS (Oracle or Sybase IQ) optimized for VLDB Distributed data storage structure split by source (domain/technology/vendor/version) and location (region) Designed for parallel processing Historical class-object-relation model for all system entities Scalable for network growth and regional splits/merges Secure data storage Flexible for customization and extension Embedded programming language for data access and modification October 2010 8 TPMS: System Description
9. Application layer Multi-threaded access to DB for parallel processing Provide open integration interface (Web-services, OSS/J, SNMP) Events generation Data aggregation, correlation and profiling Scheduled report generation Store and share generated KPIs and reports Threshold actions (alarms, notifications, etc.) Extendable with optional modules Optional clustered architecture and redundancy Automatic health-check reporting October 2010 9 TPMS: System Description
10. Presentation layer (User interface) Rich web-based user interface Report and KPI designer/browser for end-users without knowledge of SQL Dashboards and real-time reports Ad-hoc reporting with interactivity and drill-up, drill-down and drill-same capabilities Object-based and domain-based security Export report results to CSV, XML, PDF, etc. Provide an administrative UI for all system components October 2010 10 TPMS: System Description
12. Data Collection and Parsing Collect data using FTP, SNMP, CORBA, X.25, SQL, custom scripts Validate data Dump, validate and filter data Normalize, aggregate, accumulate and enrich data October 2010 12 TPMS: System Description
13. Data Loading & Validation Load parsed data into the DB Validate data gaps and data re-loads Transform and normalize late data Initiate data processing and KPI calculation mechanisms October 2010 13 TPMS: System Description
14. Data storage Keep RAW and aggregated performance data and KPIs, network hierarchy, KPI and report templates Distributed data storage structure split by source (domain/technology/vendor/version) and location (region) 1 data context = 1 DB instance or schema or database Optimized for parallel processing Designed for very large volumes of data with unstable structure October 2010 14 TPMS: System Description
15. Data abstraction Provide access to data in different contexts for presentation layer components making the data location-independent. Automatically locates requested data, builds parallelized queries and retrieves collected results. Correctly retrieves data in case of context unavailability October 2010 15 TPMS: System Description
16. KPI engine Store KPI/PI hierarchy for root-cause analysis Create KPIs by template Calculate KPIs as user-defined formulas or scripts (for complex KPIs) Aggregate KPIs by time and hierarchy Keep history of changes of KPI definitions Create personal and ad-hoc KPIs October 2010 16 TPMS: System Description
17. Report engine Store reports hierarchy Create reports by template Create batch reports or report chains Create master-detail reports Create personal and ad-hoc reports Calculate reports by request, scheduler, event Support time zones in calculations. Report may be calculated for local or central time zone Save pre-calculated report results for review and investigation without need of recalculation Save report results as XML, CSV, PDF, XLS, etc. Keep history of report definition changes October 2010 17 TPMS: System Description
18. Inventory Keep hierarchy of network elements (NE) Manage a class-object model Support vendor-specific and vendor-neutral hierarchies Keep history of changes of network hierarchy Manage virtual and logical network elements and groups (like region or data-center) Automatically discover network elements Group NEs by properties (like number of ports) October 2010 18 TPMS: System Description
19. Security engine Manage users, roles and domains Allow user access to the system functions or objects (NEs, KPIs, Reports) Provide a Single-Sign-On to the system Can be integrated with LDAP, AD, RADIUS, etc. for user authentication and authorization Log all user activities October 2010 19 TPMS: System Description
20. Alarm engine Automatically calculate KPI thresholds with minimal latency Send threshold alarms to Fault/Event Management Systems Alarms with conditions (alarm is raised in case of 2 or more threshold crosses during 1 hour) Threshold zones for different alarm severities Time-dependent thresholds Automatically clear the alarm in FM system in case of return to normal operation October 2010 20 TPMS: System Description
21. System administration System is managed from a single user interface as well as from the command line Allow system administrator to manage: Contexts System security Data in DB System components October 2010 21 TPMS: System Description
23. First steps As a first step the Performance Monitoring core functions shall be done: Data Collection and Parsing, Data aggregation and normalization, KPI engine, Reporting (tables and charts) Components to be done first: DB, Report viewer, Report designer, KPI editor, Inventory, Scheduler, User GUI October 2010 23 TPMS: System Description
24. Next steps Following Performance Management functions and components shall be added later: GIS, Alarm engine, Northbound interface, Administration GUI, Collection and parsing visual designer, OLAP, Profiler, Decision Support System, Forecast (What-If), Root-cause analysis October 2010 24 TPMS: System Description
25. Thank you. October 2010 25 TPMS: System Description PavelLechenko pavel.lechenko@hpcms.ru October 2010 This document is licensed under CC BY.