The document discusses a framework for managing mobile quality of service (QoS) using complex event processing, highlighting the importance of user experience and real-time monitoring in mobile broadband services. It details the implementation of QoS management through device-based measurements, network monitoring, and the potential for Wi-Fi offloading to alleviate congestion. Future work includes predicting congestion patterns and improving interactions with other infrastructure systems while ensuring data privacy.

1. <Insert Picture Here>
Mobile QoS Management using Complex
Event Processing
Distributed Event-Based Systems (DEBS) 2013
Conference
Mauricio Arango
July 1, 2013

2. 2
Mobile Quality of Service Management
• This work focuses on quality of mobile broadband data
services
• Key performance indicators (KPIs) include:
– Transaction latency
– Input and output throughput
• Involves monitoring and detection & resolution of
performance issues
Network
Data
Centers
Data
Centers
QoS
Management
System

3. 3
Mobile Network Quality of Service
User experience-driven trend
• Success of mobile applications largely depends on performance
perceived by end-users
• Involves device-based monitoring vs. network-only monitoring
• Very large scale of monitoring and analysis – from from 10s of
thousands network elements to 10s of millions devices
• Mobile networks are large-scale sensor networks – mobile
devices are sensors
Networ
k
Data
Centers
Data
Centers
QoS
Management
System
Networ
k
Data
Centers
Data
Centers
QoS
Management
System

4. 4
Mobile Network Quality of Service
Near real-time trend
• High end-user performance requires end-to-end QoS
monitoring and management, involving:
– Device-based apps
– Network
– Data center/cloud-based apps
• Mobile networks are massive dynamic systems –
constantly reconfiguring
– Requires rapid response to dynamic user context, application
context, network conditions
• Proactive identification and correction of problems
impacting user experience

5. 5
WiFi Offload for QoS Management
• Switch devices from mobile networks to carrier-based WiFi
networks
• Mitigate or avoid congestion
GW3G Access
AP
GW4G/LTE
AP
GWWiFi
AP
Core
Network
Internet
QoS
Management
with WiFi
Offload
Network
Provisioning &
Management
Subscriber
Information
Event & control flows

6. 6
CEP-based QoS Management with WiFi Offload
Application description
• Input events:
– Device-based measurements
• Transaction latency
• Bandwidth capacity usage per transaction
– Data center transaction latency measurements
• Detect:
– Devices experiencing high-network latency (QoS problem)
• Infer:
– Congested base stations (QoS problem)
– Device speed (mobility state)
• Actions - for every device:
– If device network latency is high AND
– Device is in a congested base station AND
– Device is static THEN
• Offload to closest available WiFi hotspot (QoS solution)

7. 7
Implementation Architecture
Oracle OEP
MASON multi-
agent simulation
toolkit
Oracle WLS Oracle
MapViewer

8. 8
CEP Application – QoS Management
Event Processing Network
QoS problem detection
– Qos solution: WiFi
offload
Input streams merge and
enrichment

9. 9
Input stream correlation and enrichment
Join device stream with
mobile base station table
Join device stream
with data center
stream
Derive
device
speed
from
location
stream

10. 10
Device speed calculation from location stream
• Partition stream, one stream per
device
• Determine device’s mobility
state: static or moving

11. 11
QoS Problem Detection & QoS Solution: WiFi
Offload

12. 12
Geospatial processing and CEP
• Mobile QoS management is a location-based
application
• Processing highly simplified with use of a geospatial
library integrated with CEP platform
• Oracle Event Processing library
– Input events and internal events mapped to a geospatial grid
– Simplified definition of queries involving location, eg:
• Nearest neighbor
• Distance
• Inclusion within boundary

13. 13
Map-based Web Application
High-latency and
congestion rendering
WiFi offload rendering

14. 14
Future work
• Predictions – mobile network infrastructure is
becoming more more reconfigurable; predict
congestion & react on time:
– Identify boundaries around high density areas – city, centers,
airports, stadiums, etc and obtain early predictions of the load
that is heading towards them.
– Identification of the speed, intensity, direction of load
movement – location and time to congestion
• Map location of points where WiFi offload would be
required
– For carriers planning to deploy WiFi offload, measure events
where there were congestion/service issues that could be
solved by offload.

15. 15
Future work – cont.
• Interactions with other infrastructure systems, eg.
transportation, power utilities
– Receive key events from them
– Provide key events to them – congestion near airport,
congestion in part of city
• Interactions with services systems and the
environment:
– Weather
– Sports
– Media programming
• Privacy is a fundamental issue:
– All of above can be and should done with anonymized data

16. 16
Key Messages
• Event processing and CEP platforms simplify
development of real-time monitoring and
management systems for mobile communications,
requiring:
– Monitoring and tracking location context of very large
numbers of moving and reconfigurable components
– Very high scalability
• Geospatial processing is a key requirement for CEP
platforms supporting location-based applications
• Mobile monitoring intersects with and can benefit
other infrastructure systems