VoLTE Performance Analysis

VoLTE Performance Analysis
An independent benchmark study of AT&T’s VoLTE
network (plus some other interesting info)
November, 2014
Test environment
Use of the ATGW
AT&T Network – Minneapolis-St. Paul, MN
(Alcatel-Lucent Infrastructure)
TEST EQUIPMENT
ITEM
Spirent’s Nomad User Experience Analytics
System
MOS, Call setup time, Call reliability
Spirent’s Quantum Battery Life Measurement
System
Power consumption, Current drain, Implied battery life
Qualcomm’s QXDM
Resource utilization, and capture RF parameters (Layer 1
measurements & PPP messages)
Accuver XCAP software and Xceed
Technologies’ WindCatcher software
Post-processing & analysis of the QXDM logs
VoLTE’s enabling technology
VOICE QUALITY
• Smartphone: natively integrated VoLTE
client and AMR-WB codec
• QCI=1 voice bearer invoked by P-CSCF via
PCRF
• QCI=1 policy enforcement across
Smartphone, LTE RAN, mobile backhaul and
S/P-GW
BATTERY LIFETIME
• Smartphone: natively
integrated VoLTE client
• E2E VoLTE: RoHC, DTX /
DRX, fewer resource blocks
& TTI, less time spent
transmitting & receiving,
reduced transmit power, etc.
VoLTE versus SKYPE: RESOURCE UTILIZATION
• Smartphone’s VoLTE interface to LTE RAN
• Robust Header Compression (RoHC)
• Discontinuous Transmission/Reception (DTX/DRX)
CALL RELIABILITY
• VoLTE network performance
• E2E VoLTE design, for
regular calls and eSRVCC
handovers
• TTI-Bundling, for cell edge
performance
CALL SETUP TIME
• Result of using VoIP instead
of 3G UMTS (CSFB), avoids
changing radios and alerting
in 3G
VoLTE and Skype over LTE MOS – Stationary with Heavy
to moderate Loading
PROCEDURES:
VoLTE MOS Test Results – IKEA, June 17th 1940 Hours
• POLQA algorithm with male
& female voices
• Stationary testing in IKEA
parking lot (near airport &
mall)
• 4 sets of tests with ~10
MOS measurements per test
OBSERVATIONS:
• Highly repeatable
• VoLTE does better than
Skype
• Less variability with VoLTE
results
Skype Voice over LTE MOS Test Results – IKEA, June 17th 2015 Hours
“Based on the POLQA
algorithm there is a clear
indication that a typical
subscriber would notice
the difference in the voice
quality between VoLTE
and Skype.”
Skype over LTE – Loading and Multi-Apps, II
Skype Voice over LTE MOS Test Results with Multiple
Applications doing Network Transactions, including on the
Test Phone
Skype Voice over LTE MOS Test Results with ONLY Angry
Birds Downloading on a Separate Phone
PROCEDURES:
•We ran other data applications in
the background on the same phone
OBSERVATIONS:
• Skype Voice could not support a
separate application transferring
data
• Downloading “Angry Birds” from
Google Play consistently caused
Skype Voice [running on different
phones] to crash
• All results shown on this slide
were repeatable
“And while other network
traffic or background traffic
downloading on a smartphone
could bring Skype Voice to its
knees, there was no indication
that it impacted the VoLTE
call.”
NETWORK RESOURCE UTILIZATION:
resource blocks and transmission time intervals (TTI)
Summary of Resource Block Requirements – VoLTE versus
Skype Voice over LTE
9.4x
6.1x
Summary of TTI Requirements – VoLTE versus Skype Voice
over LTE
3.3x
3.0x
OBSERVATIONS:
• Skype required substantially more
network resources than VoLTE for
the same voice activity patterns
 9.4x more DL RBs
 6.1x more UL RBs
 4.5x more DL TTIs
 3.9x more UL TTIs
 downlink and uplink RBs
• Also impacted battery life
(discussed in a bit)
• Increasing the voice activity
pattern will impact the results
shown here
Network resource utilization - DTX/DRX, RoHC and
PDCCH Grants
The Impact of DTX and DRX on VoLTE and Skype
Voice
The Impact of RoHC on VoLTE and Skype Voice
over LTE
3.0x
3.3x
The Impact of VoLTE and Skype Voice over LTE on
PDCCH Signaling Messages without SPS
Observations
 All three figures show results which
strongly favor
VoLTE / QCI=1 applications
4.9x
 OTT applications generally do not support
QCI=1 features
 AT&T had not implemented short DRX and
SPS so the actual benefits of VoLTE are
understated.
VoLTE and CSFB call setup times
Circuit Switched Fallback Times
PROCEDURES:
 Results based on ~150 mobile
to landline termination calls –
tested while driving around
Twin Cities
OBSERVATIONS:
 VoLTE setup time was ~twice as
fast as CSFB with slightly lower
reliability
 Mobile to mobile calls will likely
increase CSFB call setup times
and shorten VoLTE setup times
VoLTE Call Setup Times
3G Call
Reliability
VoLTE Call
Reliability
Network Resource Utilization - throughput
VoLTE Drive Test Downlink MAC Layer Throughput versus
Time
Skype Voice over LTE Drive Test Downlink MAC Layer
Throughput versus Time
PROCEDURES:
• Used QXDM to capture
chipset diagnostic information
during drive test
• Relatively low voice activity
level due to test procedure,
but identical for all
OBSERVATIONS:
• ~13x higher throughput on
Skype
• No evidence of voice activity
pattern during Skype call
• VoLTE supports RoHC and
DTX/DRX
• Short DRX and SPS are
coming
• Skype also supports DTX /
DRX, but takes less
advantage of it
Estimated Battery Life – VoLTE, Skype Voice and 3G CS
Relative Differences
Estimated Battery Life – Near the Serving Cell
PROCEDURES:
• Tested while stationary at various locations
• Selected sites based on RF conditions reported
by phone
• Used pre-recorded message to simulate a
conversation
OBSERVATIONS:
• Battery life was 56-89% longer with VoLTE
than with Skype
• 3G CS Voice resulted in the longest battery life
• The use of short DRX and/or the VoLTE client
on the modem should help VoLTE in the future
General observations
VOICE QUALITY
NETWORK RESOURCE UTILIZATION
VoLTE MOS (~3.8) was
Compared to Skype, VoLTE used
• Significantly higher than 3G CS voice (NB)
• Measurably higher than Skype (~3.4)
•
•
•
•
•
•
•
Network loading or concurrent
data transactions on the test
phone
• Brought Skype to its knees
• Had no impact on VoLTE
BATTERY LIFE
10x less bandwidth
7x fewer Resource Blocks
5x fewer TTI
6x fewer PDCCH Grants
4x fewer transmissions
89% less signaling overhead
2.3dB less uplink power from UE to tower
CALL RELIABILITY
With VoLTE, the battery
life was
Compared with 3G CS
Voice, VoLTE was
• 56-94% longer than Skype
• 11-17% less than 3G UMTS
• 96% reliable versus 99%
reliable
• A seemingly less dense LTE
RAN primarily resulted in the
differences
Affecting it were
• LTE Coverage
• VoLTE client on A/P
• Short DRX not yet enabled
eSRVCC worked to 3G
CALL SETUP TIME
VoLTE call setup time
was nearly twice as
fast as CSFB
• 2.9 versus 5 seconds
CSFB phones didn’t
always return to LTE
mode after the call
WRAP-UP AND NEXT STEPS
•
VoLTE lived up to our expectations, even in its early infancy
•
The commercial launch of additional features and network optimization /
improved coverage will further strengthen its competitive advantage
•
Additional VoLTE work is currently underway, including
– Network interoperability and roaming
– VoLTE client performance and interoperability
– Other OTT voice applications
– Video benchmarking (IR.94 versus Skype, FaceTime, etc.)
– Impact of 3G WB-AMR and higher VoLTE codec rates
– Impact of additional VoLTE-related features (short DTX / DRX, SPS,
etc.)
– RCS
Initial test results for Tmobile VoLTE using the
23.85 kbps AMR-WB codec show a MOS of 4.1
under similar non-mobile conditions. *NOTE: Small
sample size using Anite Nemo