celliq: real-time cellular network analytics at scale
TRANSCRIPT
CELLIQ: REAL-TIME CELLULAR NETWORK ANALYTICS AT SCALE Anand Padmanabha Iyer, Li Erran Li, Ion Stoica
Presenter: Chenqi Zhou, Yuhan Xu
WHY IMPORTANT?Cellular network operators collect large volume
of information about the network and user traffic
Provide crucial insights in network planning (e.g., deployment of new base stations), network operation (e.g., improving utilization
of limited radio resources by interference coordination), etc
Ensure satisfactory end-user experience
analysis
MAIN CONTRIBUTION
Proposed CellIQ—a system for real-timecellular network analytics that supports rich analysis tasks efficiently by leveraging domain-specific optimizations
CURRENT WORK
Periodically generate thousands of predefined reports available in a
dashboard
Handoffs at network elements
Radio resource usage
Associated base
stations
State-of-the-art Cellular Analytics Systems
Timely and more sophisticated analyses
l Detecting and monitoring spatio-temporal hotspots
l Tracking popular handoff sequences with abnormal failure rate
Improvement:
CELLIQlKey insight: cellular network data is naturally
represented as a time-evolving graphlLeverages domain specific characteristics of cellular
networks—its spatial and temporal locality—to achieve efficient analysis
lCapable of running sophisticated tasks including detection and tracking of spatial and temporal traffic hotspots, and popular handoff sequences with abnormal failures
STRUCTURElBackground
lChallenges & Solution Overview
lCellIQ System
lCellIQ Realization and Optimization
lCellIQ Performance and Evaluation
lConclusion and Future work
BACKGROUND
Data collected:• Bearer and Signaling
Records • TCP Flow Records • Network Element
Records
LTE Architecture
CHALLENGESThree broad tasksüContinuous monitoring of connections and entities üReal time detection of spatial and temporal patterns üReal time troubleshooting to identify root causes
Performing these tasks in real-time can be very challenging!
THREE OPERATIONS• Sliding Window
Operations − Persistent hotspot tracking per
sliding window
• Time Window Operations − Popular handoff sequence
tracking per time window
• Spatial Operations − Top traffic gradients tracking
SOLUTIONLeveraging cellular specific optimizations exploiting spatial and temporal locality:• Data placement • Radius based message broadcast • Spatial aggregation • Differential graph updates • Incremental graph updates
CELLIQ SYSTEM-GRAPH PARTITIONING
Graph computation frameworks rely on partitioningto minimize communication & balance computation
• vertex-cut strategy: avoid replicating edges• geo-partitioning technique: place edges to preserve spatial locality
CELLIQ REALIZATION AND OPTIMIZATION
• 1. GeoGraph Computational Model
• 2. Gstream Computational Model
• 3. Co-partition Component Graphs
• 4. Indices and Routing Tables
GSTREAM APITRACKING PERSISTENT HOTSPOTS
GOAL: DETECT AND TRACK GROUPS OF BASE STATIONS WITH HIGH TRAFFIC VOLUME
CONCLUSION AND FUTUREWORK
• Conclusion:CellIQ is a cellular network analytics system that uses domain-specific optimizations to achieve 2x to 5x improvements
• Future Work:• Using techniques in CellIQ to perform root-cause
analysis on operational LTE Networks• Generalized streaming graph analysis techniques