VoLTE E2E Optimization

NPO Seminar Bangkok: NPO Japan - Docomo and Softbank PoC

• Niklas Gustavsson
• 15-06-2016

1 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Agenda

• Introduction
• SIP Registration Analysis
• Call Setup time analysis
• VoLTE MOS and Mute call analysis
• SIP Call Setup - INVITE & Cause code Analysis
• SIP Call Termination - BYE Cause code Analysis
• Feature and Contact Parameter analysis
• Summary and Recommendations

2 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Probe Setup
Overview

Docomo Softbank
MME MME

S1-C S1-C
eNB eNB
S11 S11
SGW IMS/Core IMS/Core
SGW
S1-U S1-U

Probe
Probe

Docomo: Softbank:
• No NW filtering – capture all • No S1-C tapping in Softbank
• Differentiated services (DSCP) HW capturing filter applied

3 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Introduction
Docomo Softbank

Captured S1-U, S1-MME, S11, S6a S1-U/S5/S8 and S11 – FlexiNG (SGW)
interfaces
Captured 2.6 TB – LTE and VoLTE data 38GB – Only VoLTE
amount of
Docomo:
data VoLTE Traffic in Uplink
25%
Capture date 2015-08-26 (11.50-13.00) 2015-10-14 (11.50-14.00) VoLTE Traffic in Downlink
3%

Captured calls 1610 5000

Average Call 32s 2 min 9s

duration
Average MOS 4.2 4.2

EXFO Travelhawk Pro

4 © Nokia Solutions and Networks 2014
SIP Registration Analysis

5 © Nokia Solutions and Networks 2014

 UE MME/SAE/PGW S-CSCF
SIP Registration
Signalling flow Attach, Default Bearer setup, NAS Authentication, Integrity
Protection 1
Activate EPS Bearer (QCI5) and DRB
1. UE attaches to the network, the PGW assigns IP
Activate EPS Bearer Accept (QCI5) and DRB
address and identifies P-CSCF to the terminal.
2. UE sends Registration Request to S-CSCF (via P- Store P-CSCF IP
Address
CSCF) in REGISTER message. Extract user public
3. S-CSCF performs registration procedures with HSS and identity from ISIM

acquires user authentication information and sends UE Allocate client and

a challenge in 401 Unauthorized message.

server ports
2
REGISTER
4. UE calculates the response and sends it to S-CSCF in Registration procedures
REGISTER message.
3
with HSS

5. After the authentication has succeeded the S-CSCF 401 Unauthorized:

downloads and stores service control information from
Verify AUTN &
HSS and S-CSCF notifies the terminal about completed calculate RES

registration with 200 OK message. 4 REGISTER

Download service control

information from HSS

5
200 OK,

6 © Nokia Solutions and Networks 2014

Docomo: SIP Registration
Expected re-registration is after 3600 seconds
• IMS is pushing expires timer in 200 OK
response to REGISTER message, IMS is
setting this to 3600 (Expires header)
• In the initial REGISTER request from UE
this can be a higher but at Docomo it is
3600 as well.
• SIP Re-Registration can happen before
the timer expires.
• The IMS and the UE can trigger re-
registrations some scenarios (Ref 3GPP TS
24.229), explaining some UE have 2-3
registrations in the 70 min.

Conclusion: Nothing alarming with the number of SIP registration per UE

7 © Nokia Solutions and Networks 2014

Docomo: SIP User-Agent info in Registration requests
80% of SIP Register request do not provide any info about the User-Agent

• Samsung and iPhone is

providing this information
sometimes, but not always.

• The User-Agent information is

sometimes provided in SIP
INVITE message instead.

• It is good practice to provide

the User-Agent header in the
SIP registration in an IMS
context as IMS can take actions
based on this header.

8 © Nokia Solutions and Networks 2014

Softbank: SIP Registrations
Average SIP Registrations is 1.3 per 15 minutes ~ 5 per hour
• Registration expiry header
timer is set to 2390s (40
min)

• Often the UA is
challenged with 401 and a
challenge string as
response Status. This
means another
REGISTRATION message is
sent with a calculated
nonce, so in total 2 SIP
Registrations messages are
sent.

• Therefore around 3 SIP

Registrations per hour are
expected in average per UA

9 © Nokia Solutions and Networks 2014

Softbank: SIP Registrations per UA
Older iOS versions and Samsung UAs are causing above average Registrations

• Most common UA is
iOS/9.02, end-users are
updating regularly to
Samsung latest iOS from Apple

iOS 8 • Very few Samsung

iOS 9 mobiles seen
No Agent
• Less than 10% of UA is
Mainly Sony not providing User-Agent
and Sharp
information, they are
mainly Sony and Sharp.

• iOS 8 and Samsung IMS

client generate
significantly more
requests
10 © Nokia Solutions and Networks 2014
Softbank: SIP Registration Success ratio per UA
99.7% in overall – which is good. Some iOS versions such as 8.4 have poorer ratio but are not
common among the subscribers

11 © Nokia Solutions and Networks 2014

Softbank: Reason for iOS8.4 iPhone low Registration Success rate
is missing “Require” and “Proxy-Require” parameters.

Normal

Missing Require: sec-agree • 3GPP IPSec is mandatory in IMS

• Even if NULL cipher is used

• Reason can be malfunction SIP client

in certain Apple UE
• In total 42 such cases are seen
• No re-attempt with sec-agree field is
No “Require” “Proxy-
Require” parameters
seen
12 © Nokia Solutions and Networks 2014
Softbank: Other reason for SIP Registration failure with iOS9.0.2 (2)
the reason is changing UE IPv6 address frequently
Time Source(UE) Destination(CSCF) Protocol CSeq
45:49.1 2407:5300:1844:59d1:14a4:22d9:5d2:6943 2407:5300:2000:5024::6 GTP <SIP> 564 REGISTER
51:34.1 2407:5300:1844:59d1:14a4:22d9:5d2:6943 2407:5300:2000:5024::6 GTP <SIP> 565 REGISTER
51:40.4 2407:5300:1820:8b9b:62:502c:f998:1713 2407:5300:2000:5021::4 GTP <SIP> 567 REGISTER
56:34.8 2407:5300:1820:8b9b:62:502c:f998:1713 2407:5300:2000:5021::4 GTP <SIP> 568 REGISTER
56:41.3 2407:5300:1800:b7b7:185c:4cd0:53cc:1d4 2407:5300:2000:5021::12 GTP <SIP> 570 REGISTER
57:01.4 2407:5300:1800:b7b7:185c:4cd0:53cc:1d4 2407:5300:2000:5021::12 GTP <SIP> 571 REGISTER
57:13.3 2407:5300:1860:64c2:86a:e965:375d:695d 2407:5300:2000:5023::9 GTP <SIP> 573 REGISTER
58:38.4 2407:5300:1860:64c2:86a:e965:375d:695d 2407:5300:2000:5023::9 GTP <SIP> 574 REGISTER
58:46.0 2407:5300:1800:df11:1098:a6b1:c5ae:f646 2407:5300:2000:5021::8 GTP <SIP> 576 REGISTER
58:55.6 2407:5300:1800:df11:1098:a6b1:c5ae:f646 2407:5300:2000:5021::8 GTP <SIP> 577 REGISTER
59:09.5 2407:5300:1842:6378:1822:b7b8:245:a410 2407:5300:2000:5024::b GTP <SIP> 579 REGISTER
59:13.6 2407:5300:1842:6378:1822:b7b8:245:a410 2407:5300:2000:5024::b GTP <SIP> 580 REGISTER
59:32.4 2407:5300:1843:b8df:103e:2231:64dd:3c74 2407:5300:2000:5023::d GTP <SIP> 582 REGISTER

No real need to re-request IP addresses - in some cases handover or delete

session is the reason
14 © Nokia Solutions and Networks 2014
Softbank: Recommendation & next steps
Area Recommendation Benefit
Reduce the number of SIP Understand why a new IP is allocated so often - UE battery saving
Registration more than 1 per minute in some cases. IMS load reducing
Using static IPv6 address instead of dynamic
should reduce the number of Registration
Optimize the Expires header(2390s) unless there is
a specific reason.
(RFC 3261 mention 3600s as default for expires)
Improve the SIP Registration Investigate Samsung IMS client. End user experience
Success ratio
Start investigation on UE with iOS 8.4 – Wrong SIP
Registration request – missing sec-agree field. This
means such UEs cannot Register and not initiate or
receive VoLTE call.
Investigate UEs (with iOS 9.0.2) that are sending a
flood of SIP Registrations.

15 © Nokia Solutions and Networks 2014

Call Setup time analysis

16 © Nokia Solutions and Networks 2014

Call setup times Analysis
Definition of call setup time for MO and MT

UE A SGW SGW UE B

INVITE
Paging
RACH
INVITE
MO Call MT Call
Setup Time Setup Time
180 Ringing

180 Ringing

200 OK
200 OK

17 © Nokia Solutions and Networks 2014

 Originating UE IMS Network Terminating UE
SIP Call - Reference
VoLTE to VoLTE Initiate call

INVITE

100 Trying
INVITE

EPS Bearer Activation for QCI1 and Audio Video Path Setup

183 Session Progress

PRACK PRACK

200 OK 200 OK

UPDATE UPDATE

200 OK 200 OK
Ringing

180 Ringing 180 Ringing

PRACK PRACK

200 OK (prack) 200 OK (prack)

Answer

200 OK (invite) 200 OK (invite)

ACK ACK

Voice or Video Session

BYE BYE

ACK ACK

18 © Nokia Solutions and Networks 2014 Note: ACK for BYE is 200 OK instead in most cases
Docomo: Call setup times
Some calls take a long time to setup

• Some calls take a long time to setup –

SIP Call Setup time 11.50 - 13.00 around 10% takes longer than 5s which
2000 100.00%

1800 90.00%
is not normal
1600 80.00%

1400 70.00%
• The graph is showing both MO and
1200 60.00%
Frequency

1000
10% of all calls takes > 5s to setup 50.00%
MT combined
Frequency
800 40.00%
Cumulative %
600 30.00%
• A Check was done for 2 INVITE
messages and 2 RINGING – that is a
400 20.00%

200 10.00%

0 0.00% call was done in the same SGW. This

0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.5
9.0
9.5
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
19.0
20.0
More
would indicate the paging time. No such
Seconds
cases were found. Might be that it is
considered a duplicate packet and
Long call setup time filtered out

19 © Nokia Solutions and Networks 2014

Docomo: Paging
Control plane statistics for paging – QCI 1 bearer
UE eNB MME

Paging Paging

Paging response
Service Request time

• All paging messages are responded within 1.5s

• Reason for long setup time is not paging Paging response time:
Definition: Paging to Service Request message

20 © Nokia Solutions and Networks 2014

Docomo: SIP messages
Caused by IMS NE or UE B-Party
Delayed Ringing

Abnormal call setup time

15s

13.8s

Normal call setup time

1.8s

21 © Nokia Solutions and Networks 2014

Softbank: Call Setup Time – All calls
Two distinct peaks can be seen.
•Average Call
setup time: 2.6 s
• Max Call setup
time: 13.7 s

• Second peak
includes paging
17% of all calls takes > 5s to setup attempts: the
time difference
to first peak is
around 4s which
probably is the
setting of timer
T3413

22 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Softbank: Call Setup Time – Originating & Terminating calls
MO has 2 peaks.

• Paging time is not included in the setup • The call setup time for originating calls includes
time and represents the call setup time for paging before the SIP INVITE message can be
pure SIP signaling between A and B party sent out.
• There is first peak around 2.5s which probably
means the UE is reached by first page
• A majority of all calls needs a second page!

23 © Nokia Solutions and Networks 2014

 Softbank: Long Call Setup case(2) [11.5s]
First Session Progress delayed 6s.

6s

• The B-Party is another Mobile (UE)

iPhone. The reason for the delay
might be due to long paging

• Investigate cell & radio environment

in paged cell.

24 © Nokia Solutions and Networks 2014

Softbank: Call Setup time around 2nd peak [5.5s]
Same reason as long call setup cases – Ringing or 1st Session Progress delayed.
Ringing delayed Session Progress delayed

-> “Second Peak” delay not only linked to Paging

25 © Nokia Solutions and Networks 2014
Softbank: Recommendation & next steps
Area Recommendation Benefit
Paging time Optimize paging mechanism( stepwise paging. Etc.) End user experience
in order to speed up paging i.e. whole TAI paged
immediately.
Revise the setting of timer T3413 (paging retry
timer)
Call setup (excluding paging) Check IMS loading - delay reason could be End user experience
slow/loaded IMS NE – especially delayed RINGING
message
The goal should be to have 95% of call setup times
less than 5s – Now only 83% for all calls. For
originating calls only 60%.
To investigate further by probing S1-MME, Mw, Cx
and ISC/Sh interfaces as next steps.

26 © Nokia Solutions and Networks 2014

VoLTE MOS and mute calls analysis

27 © Nokia Solutions and Networks 2014

Docomo: Low MOS field analysis: Summary
51 cells where min. MOS value was less than 3 were listed up.
- Min. MOS value is the minimum value of MOS per one second within the call duration.
Out of the 51 cells, walking test was performed for 21.
- Considering the access method, Nokia selected cells where walking test was
performed.
Out of the 18 cells where measurement was possible, low MOS issue was confirmed in
17 cells.

Low MOS Measurement No low MOS Total

found impossible
17 cells 3 cells 1 cell 21 cells
95% (N/A) 5% 100%
28 © Nokia Solutions and Networks 2014
 Docomo: Low MOS field analysis: Result summary
• Events that are suspected as causes of MOS degradation found by the walking test.
Probe MOS degradation location found by the walking test
Indoor Measu
MOS weak RRC
Sector name weak HO remen
min. covera interfere disconne Comment
Min. coverag failure t date
value ge nce ction
MOS e
ＣＶ神田駅第二1 0.88 1.78 ✔ ✔ ✔ アセンド神田ビル 1Fロビーで劣化 9/10
◆ＣＶ幡ヶ谷2 1.04 1.16 ✔ ✔ ✔ ✔ ✔ 幡ヶ谷駅周辺店内及び局近傍で劣化 9/11
◆ＣＶ新橋北＃１1 1.35 1.62 屋外でのMOS劣化は殆ど見られない 9/10
ＣＶ南青山二丁目ＲＫ1 0.89 1.14 ✔ ✔ ✔ 北青山１丁目アパート付近で劣化 9/11
◆ＣＶ上野駅南ＲＫ1 1.75 1.07 ✔ ✔ 上野駅２F歩道上で劣化、切断 9/15
ＣＶ桜田門1 1.11 1.00 ✔ ✔ ✔ ✔ 地下鉄二重橋駅出入り口階段で劣化 9/8
品川ＴＷＩＮＳ＃２3 1.26 1.30 ✔ ✔ 屋内局とのネイバー抜けの可能性有 9/15
ＣＶ西台第三6 2.04 2.12 屋外での劣化は特に見られない 9/14
ＣＶ板橋第二＃２2 1.76 1.51 ✔ ✔ 板橋市場入口交差点で劣化 9/14
ＣＶ高島平団地ＲＫ第二1 1.12 1.00 ✔ ✔ 東武ストア店内で弱電界による切断 9/14
ＣＶ東上野四丁目ＲＫ第四1 1.4 2.42 屋外での劣化は特に見られない 9/15
ＳＴ東京交通会館2 0.89 1.16 ✔ ✔ ✔ 地下駐車場の階段で劣化 9/10
ＣＶ代々木二丁目ＲＫ第三3 2.72 1.16 ✔ 周波数間HOの失敗で劣化 9/11
ＣＶ代々木三丁目ＲＫ第二2 1.94 2.16 ✔ 首都高速新宿PA近傍で劣化 9/11
ＣＶ赤坂九丁目ＲＫ第二2 1.04 1.00 ✔ ✔ ✔ ✔ ミッドタウンン地下入口で劣化 9/11
ＣＶ志村三丁目ＲＫ第三1 1.72 1.77 ✔ ✔ 2GでUL BLER上昇によるパケットロス増加あり 9/14
ＣＶ外神田四丁目ＲＫ第二1 1.34 1.44 ✔ ✔ 地下鉄末広町駅出入り口階段で劣化 9/10
ＳＴＪＲ目黒駅第四＃１2 0.85 1.13 ✔ ✔ JR連絡階段で劣化。ネイバー追加後に劣化減少 9/8
29 © Nokia Solutions and Networks 2014
Softbank: MOS
Average MOS [4.18] is good but there is some calls which Average Minimum MOS is less than 2.

Average Min MOS: 3.56 Average MOS: 4.18

Note: MOS is calculated every second by the probe

30 © Nokia Solutions and Networks 2014

Softbank: Normal VoLTE call
One RTP stream high bit rate at the time – As one party is talking other party is listening

A-Talking B-Talking

31 © Nokia Solutions and Networks 2014

 Softbank: Mute RTP streams examples
One-way audio
One-way audio example(1) One-way audio example(2)

• Call is maintained but B-party has no audio • RTP stream mutes in one direction beginning of
flow, RTP stream stops in one direction after the call for 5 seconds.
10s. • Cell-ID: 44020303b271xxxx
• Cell-ID: 44020c088bf8xxxx

32 © Nokia Solutions and Networks 2014

Recommendation & next steps
Nokia has a method how to identify mute calls from RTP stream characteristics (low MOS, low
throughput and high packet loss).

Recommendation Benefit
Softbank: One-way RTP or Do measurements 1 week for defined eNB End user experience
mute calls Cluster(s) in Kyushu area.
Low MOS calls Determine worst cells identify
Additionally probe at least S1-MME interface to
root causes and apply specific
gain mobility insight for those mute sessions.
capacity rules
Correlate with eNB OSS statistics for those
clusters Identify if certain UE types and
MGW having more problem
Correlate mute calls with MGW destination and UE
than others – serve as start for
types.
real root cause detection.

Docomo: The probe is good at identify cells with bad MOS – Focus drive tests where it
95% accuracy confirmed with walk test matters. Reduce costs for
drive tests.

33 © Nokia Solutions and Networks 2014

SIP Call Setup
INVITE & Cause code Analysis

34 © Nokia Solutions and Networks 2014

Softbank: SIP INVITE cause codes distribution
2% of SIP invite have error cause codes.

Comments
• 481 some reason is Cancelled call but not all, see details next slide
• 486 and 487 are usually normal “call clearance”
• 603 the B-Party do not want to answer the call
• 200 is an OK to a BUSY is not failed calls
• 380 Alternative Service is unusual Status – explained next slides
• 503 is explained next slides
35 © Nokia Solutions and Networks 2014
<Change information classification in footer>
 Softbank: Issue(1)_wrong reason sent
75% of SIP INVITE error cause code is “380 Alternative Service”.

• The 380 Alternative Service message from the

reason text means – “Emergency calls are not
support via VoLTE”
• However none are real emergency calls – many
are call to mailbox 1416 or KDDI/AU support
number 157, some other Directory numbers to
NTT.
• Recommendation:
• Provide correct routing to some Directory
services (compare with CS IN service)
• Change the misleading reason text.
End User impact:
• Investigate why some user cannot reach their
• Some user cannot reach their voice mailbox
mail box?
• Cannot reach certain IN (short) numbers.
• Misleading reason text can confuse clients if parsed
36 © Nokia Solutions and Networks 2014
Softbank: VoLTE Call Setup failure due to Handover in Progress
Failure Message Sequence - 503 Service unavailable
S-eNB T-eNB MME S-SGW T-SGW PGW

SIP:INVITE

SIP: TRYING

SIP: SESSION PROGRESS

Radio Handover in Progress

Create Bearer Request Create Bearer Request (QCI1)

E-RAB Setup Request

E-RAB Setup Response (failure) Cause: X2-Handover-Triggered

Patch Switch Request

Create Bearer Response (failure)Cause: Temporarily Rejected due to Handover Procedure in Progress

SIP: 503 Service Unavailable

37 © Nokia Solutions and Networks 2014

Softbank: VoLTE Call Setup failure due to Handover in Progress
Correct Message Sequence
S-eNB T-eNB MME S-SGW T-SGW PGW

SIP:INVITE

SIP: TRYING

SIP: SESSION PROGRESS

Radio Handover in Progress

Create Bearer Request Create Bearer Request (QCI1)

E-RAB Setup Request

E-RAB Setup Response (failure) Cause: X2-Handover-Triggered

Patch Switch Request

Retransmission
Create Bearer Response (failure) Cause: Temporarily Rejected due to Handover Procedure in Progress

Create Session Request (QCI5)

Modify Bearer Request

Modify Bearer Response

Create Session Response
Patch Switch Request Ack

Delete Access Bearer in S-eNB and Session in S-SGW

Create Bearer Request Create Bearer Request (QCI1)
SIP: PRACK

38 © Nokia Solutions and Networks 2014

Softbank: Recommendation & next steps
Area Recommendation Benefit
380 Alternative Service Investigate Mailbox and CS IN settings or IN Services working in VoLTE
compatibility in VoLTE/IMS context domain

Multiple INVITE Investigate IMS and/or SIP UA iOS 9.0.2 iPhone Avoid simultaneous call setup
between same parties

Cause “Temporarily rejected Check P-GW mobility procedures and timers End user experience: The call
due to handover procedure in (vendor dependent) cannot be setup during intra-
progress” RAT (LTE) handover

39 © Nokia Solutions and Networks 2014

SIP Call Termination
BYE Cause code Analysis

40 © Nokia Solutions and Networks 2014

Softbank: BYE cause codes distribution
It looks normal but some abnormal cases found in next slide.

• 481 Status is normal in many cases as a response to BYE if both parties hang
up at the same time – The call leg already gone for one party.
• Non normal case explained in next slide.

41 © Nokia Solutions and Networks 2014

 Saftbank: Issue(1)_Phantom Call
Call is Answered despite Cancelled by A-Party

3GPP-E-UTRAN-TDD; utran-cell-id-3gpp=44020A41E89C6668
14s
3GPP-E-UTRAN-FDD; utran-cell-id-3gpp=44020A41E4C024D5

• Call is cancelled by A-Party after 14s

• Cell-ID has changed i.e. Handover between Session Progress and Ringing
• WRONG: CANCEL is repeated multiple times

GTPC - Delete Bearer Response – Context not found (not shown in message flow)

• Wrong: The call is answered, 200 OK to INVITE

Continue…
42 © Nokia Solutions and Networks 2014
Issue(1)_Phantom Call_cont’d

3GPP-E-UTRAN-FDD; utran-cell-id-3gpp=44020244A4C1FCD3

3GPP-E-UTRAN-FDD; utran-cell-id-3gpp=44020244A4C3C35B

•Call is “answered” multiple times with 200 OK

• Call is finished after 146 seconds

43 © Nokia Solutions and Networks 2014
Softbank: Recommendation & next steps
Area Recommendation Benefit
Phantom call Further investigation needed for cancelled calls If not fixed, both parties can
during intra-RAT handover potentially not make a new call
in 2min 26 seconds until BYE is
sent by B-Party

44 © Nokia Solutions and Networks 2014

Feature and Contact Parameter analysis

45 © Nokia Solutions and Networks 2014

Feature and Contact Parameter - What are they used for?
Used by UAs to indicate their capabilities (suppl. services, SRVCC, video) to the network

Contact parameters Feature Tag

Sent during REGISTER Sent during REGISTER or INVITE
UE must reveal mandatory capabilities according The feature tags is a way to express the
to GSMA IMS Profile for Voice and SMS. capabilities.
Typical example in a VoLTE context is to express Typical example in a VoLTE context is to express
the support of SMS over IP and MMTEL the flavors of SRVCC the UE supports
(supplementary services, call forwarding, voice Example: *;+g.3gpp.srvcc -> UA supports SRVCC
mail etc) the UE supports.
Example 2: ;+g.3gpp.mid-call -> Indicates UA
Example: +g.3gpp.icsi-ref="urn%3Aurn- support for Conference Call and Call On Hold
7%3A3gpp-service.ims.icsi.mmtel“ and during SRVCC.
+g.3gpp.smsip -> this feature-tag indicates that
the device is capable of accepting SMS messages
via SIP as specified by ETSI TS 124 341. See RFC 6809 for details.

46 © Nokia Solutions and Networks 2014

Docomo: Feature Tag Comparison iPhone and Samsung
Samsung support more advanced features than iPhone

47 © Nokia Solutions and Networks 2014

Docomo: Recommendation & next steps
Conclusion

Area Recommendation Benefit

More Samsung UE are iPhone supporting more
supporting advanced SRVCC advanced features and more
features than iphone, for likely to be able to complete
example pre-alerting SRVCC. Check capability of UE HW installed base the call setup during pre-
Discuss with UA/UE vendor possibility to include alerting/alerting phase
relevant contact/feature parameters.
Contact parameter are less Avoid undetermined capability
frequent present of Supplementary Services and
SMS over IP

• All contact parameters are conforming to 3GPP, GSMA and RFC

recommendations.

48 © Nokia Solutions and Networks 2014

Softbank: No SRVCC function in some iOS 9.0.x
10% of all iPhones (iOS 9.0.x) are not providing any Feature Tag at all so SRVCC doesn’t work.

•Note: If SRVCC in altering phase is supported then “g.3gpp.srvcc-alerting” must be provided by UE see 3GPP TS 24.237

49 © Nokia Solutions and Networks 2014

Softbank: Recommendation & next steps
Area Recommendation Benefit
10% of UA 9.0.x does not Avoid end user dropping call
provide support for SRVCC during SRVCC

Check capability of UE HW installed base

Short message over IP/IMS Discuss with UA/UE vendor possibility to include Allow SMS over IP
relevant feature tag during call setup

Mid-call feature during SRVCC Maintain conference call/call

on hold during SRVCC

50 © Nokia Solutions and Networks 2014

Summary and Recommendations

51 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Summary and Recommendations - Registration
Area Issue Recommendation Benefit

SIP Registration Reduce the number of SIP Registration • Understand why a new IP is allocated UE battery saving
so often - more than 1 per minute in IMS load reducing
some cases.

• Using static IPv6 address instead of

dynamic should reduce the number
of Registration.

• Optimize the Expires header(2390s)

unless there is a specific reason.(RFC
3261 mention 3600s as default for
expires)

Improve the SIP Registration Success • Investigate Samsung IMS client. End user experience
ratio

• Start investigation on UE with iOS 8.4

– Wrong SIP Registration request –
missing sec-agree field. This means
such UEs cannot Register and not
initiate or receive VoLTE call.

52 © Nokia Solutions and Networks 2014

Summary and Recommendations – Call Setup and Tear down
Area Issue Recommendation Benefit

Call Setup and Tear Down Paging time Optimize paging mechanism( stepwise End user experience
paging. Etc.) in order to speed up paging
i.e. whole TAI paged immediately.

Revise the setting of timer T3413

(paging retry timer)

Call setup (excluding paging) Check IMS loading - delay reason could End user experience
be slow/loaded IMS NE – especially
delayed RINGING message

The goal should be to have 95% of call

setup times less than 5s – Now only 83%
for all calls. For originating calls only
60%.

To investigate further by probing S1-

MME, Mw, Cx and ISC/Sh interfaces as
next steps.

380 Alternative Service Investigate Mailbox and CS IN settings or IN Services working in VoLTE domain
compatibility in VoLTE/IMS context

Multiple INVITE Investigate IMS and/or SIP UA iOS 9.0.2 Avoid simultaneous call setup between
iPhone same parties

Cause “Temporarily rejected due to Check P-GW mobility procedures and End user experience: The call can be
handover procedure in progress” timers (vendor dependent) setup during intra-RAT (LTE) handover

53 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Summary and Recommendations – Speech Quality
Area Issue Recommendation Benefit

Speech Quality One-way RTP or mute calls Do measurements 1 week for defined eNB Cluster(s) in End user experience
Low MOS calls Kyushu area.
Determine worst cells identify root
Additionally probe at least S1-MME interface to gain causes and apply specific capacity
mobility insight for those mute sessions. rules

Correlate with eNB OSS statistics for those clusters Identify if certain UE types and
MGW having more problem than
others – serve as start for real root
Correlate mute calls with MGW destination and UE types. cause detection.

Where to focus my speech quality The probe is good at identify cells with bad MOS – 95% Focus drive tests where it matters.
drive tests accuracy confirmed with walk test Reduce costs for drive tests

UE capability 10% of iPhone 9.0.x does not Avoid end user dropping call during
provide support for SRVCC SRVCC

Short message over IP/IMS Check capability of UE HW installed base Allow SMS over IP
Discuss with UA/UE vendor possibility to include relevant
feature tag during call setup
Mid-call feature during SRVCC Maintain conference call/call on
hold during SRVCC

User Agent info missing IMS can take actions based on this
header.

54 © Nokia Solutions and Networks 2014

Some final reflections…

Call failure during intra RAT (LTE) handover during call setup and call tear down – VoLTE still immature
technology – Interaction between EPC and IMS issues.

Probe good at finding cells with low MOS score – Tell you where to focus drive tests.

Probe had problem to filter out relevant VoLTE from all capture data – Use HW filter to capture only
VoLTE traffic.

Tool refinement needed to find mute calls from RTP stream characteristics (low MOS, low throughput
and high packet loss) – Prototype exist in python…
Probe cannot analyze VoLTE mass data – Mainly design to trouble shoot single calls - external tool
needed for VoLTE mass data analysis

55 © Nokia Solutions and Networks 2014

<Change information classification in footer>
Probe comparisons

Probe Vendor Pros Cons

EXFO Good at capture/filtering Very poor support for
and export pcap Mass data analysis
Viavi Solution (former Unreliable interface Some Support for mass
JDSU) capturing keep crashing data analysis in control
plane
Empirix Good at capture and Expensive
mass data analysis
Intermediate export
format ASDR
Tektronix Very good capabilities Very expensive
for Mass data analysis
56 © Nokia Solutions and Networks 2014
<Change information classification in footer>