Testing LTE Network Performance for New Service Requirements
Nov 15, 2015
Sponsored by:
Testing LTE Network Performance for New Service Requirements
Sponsored by:
Today's Presenters
Patrick Donegan
Chief Analyst
Heavy Reading
Assaji Aluwihare
Director of Product Management,
Assurance Solutions, JDSU
Sponsored by:
Webinar Logistics
Participate in the webinar: Ask questions, share
feedback via the survey, and access the Information
panel.
Personalize your experience: Click the buttons at
the bottom of your screen to open supporting content
and user tools at your own convenience.
Technical Issues: Ask the support team for live
assistance in the ask-a-question window.Enjoy the webinar and thank you for viewing!
Sponsored by:
Why The Mobile Network Still Isnt Meeting Expectations
13%
16%
39%
45%
48%
39%
0% 20% 40% 60% 80% 100%
Outage
Degradation
All networks & services are affected
Some networks & services are affected, but not all
Only one type of network or service is affected
Source: Mobile Network Outages & Service Degradations, October 2013
Outages & Degradations
Congestion the main cause of degradations
The well-known problem of bursty traffic hasnt gone away.
User impact in the first five years of mobile broadband era.
Risk of a far greater impact with new service innovation around delay-sensitive apps like VoLTE, interactive video & gaming.
Operator business objectives heavily reliant on these new services.
Sponsored by:
Planning & Configuring The Mobile Network
Outputs Packet priortization Capacity planning Configuration settings Policy control Committed Information Rate (CIR), QoS Queues and
the Committed Burst Size (CBS) buffers.
Inputs Network Test & Measurement results. How the network has been performing.
A step change needed in outputs & inputs Phenomenal job being done More intelligent contention without congestion
Sponsored by:
Testing Performance In The Backhaul An Increasingly Heterogeneous environment
L2 & L3 protocols
Different physical layers
Multiple hops
Different architectures
Third party wholesalers
Sponsored by:
Traditional Performance Monitoring Does Not Reflect User Experience
Todays PM is synthetic and inadequate PM methods are single technology, networks are
heterogeneous
End-to-end tests often only provide a roundtrip view
Synthetic tests must be designed into the network
Samples are at fixed time slots, do not adapt to real traffic
Pre-defined test windows often miss transient issues
Traditional service management is inadequate Only uses statistical data from network elements
May only have data from synthetic test based tools
RFC 2544
Y.1564
Y.1731
TWAMP
End-to-end visibility requires correlation of huge amounts of data
Sponsored by:
Long Traffic Samples Intervals Disguise Smaller Bursts
0
50
100
150
200
250
300
Link Total Bandwidth - Downlink (Mbps)
DateTime1 (1 secs)
DateTime2 (30 secs)
DateTime3 (60 secs)
DateTime4 (300 secs)
1-15min sampling intervals causes smoothing of QoSimpacting traffic
Limits visibility of bursts
Critical for traffic engineering, shaping and policing
Sponsored by:
Lack of Segmentation Delays Trouble Resolution
Modern networks have complex aggregation schemes hub and spoke
End-to-end tests do not reveal performance at each hop
Problem isolation could take hours or even days, involving expensive dispatches
Sponsored by:
Segmented Performance Monitoring With Live Traffic
Hop-by-hop visibility into performance
Granular view of utilization finds bursts
Must meet MEF compliant standards-based testing
Sponsored by:
The Value of Segmentation and Performance Monitoring with User Traffic
Live subscriber performance metrics across routes, by circuit, DSCP value, or service
Passively monitor application performance
Measure to rate of traffic not fixed timeslots
Find network hops and QoS values with excess one-way delay or jitter in real time
Capture packets for deep analysis
Detect and correct or optimize worst hops contributing to poor E2E performance
Sponsored by:
Use Sampling Rates That Adapt to Traffic Rates
Traditional periodic sampling leaves gaps
Real-traffic based monitoring adapts the number of samples to the data rate
Visibility into true performance is preserved
Sponsored by:
High Resolution Sampling Provides Visibility into Micro Bursts
Model networks without overly averaged utilization
Find burst by seeing live traffic: 1 second metrics
100ms micro-burst analysis
Real Example: Micro-burst peak observed at 420% of network element metric
Sponsored by:
Use Case: Find Bandwidth Utilization on Key Circuits
Monitor bandwidth utilization:- Total bandwidth used- Bandwidth by DSCP class
For each path with aggregated traffic to and from eNodeBs
For groups of paths:- a group of traffic classes- a group of eNodeBs in a region- a group of eNodeB attached to a specific RNC
Aggregate groups up to regional or all network level (management dashboard)
Compare current used BW with baseline obtained from turn up tests
Understand usage trends
Save OPEX
Backhaul can be 26% of operator OPEX*
Identify opportunities to consolidate backhaul
Identify opportunities to re-engineer circuits
Save OPEX
Backhaul can be 26% of operator OPEX*
Identify opportunities to consolidate backhaul
Identify opportunities to re-engineer circuits
*Source: Amdoc 2013
Sponsored by:
Conclusion: The Benefit to Operators
Save OPEX on backhaul
Understand and optimize backhaul links for actual throughput and QoS requirements
Find misconfigured network elements
Identify, unexpected, superfluous traffic
Measure at true rate for shaping and policing (100ms vs mins)
Ensure all third party backhaul SLAs are being met
Accelerate service deployment, minimize impact on subs, resolve issues faster
Understand how new services impact existing service performance of real subscriber traffic vs synthetic traffic
Ensure quality of low-latency services like VoLTE, in real time
Avoid problems such as microbursts; measure real traffic loads, avoid under sampling with synthetic traffic
Detect QoS problems that could create costly issues such as customer dissatisfaction due to calls drops
Sponsored by:
Q&A