Double Migration
of Packet Clocks
Kenneth Hann
Principal Engineer
Artwork:Tanja Hann
November 1, 2011 1
Packet Clocks...
the first migration
Land of
Phase
Data
Com
Republic
Legacy
Land
Packet Clocks...
the first migration
Land of
Phase
Data
Com
Republic
Legacy
Land
Packet Clocks...
the first migration
Land of
Phase
Data
Com
Republic
Legacy
Land
10/25/10 5
Agenda
1. From “here”...2. Timely progress3. Supporting PTP migration
4. To Phase (eventually)
5. Conclusions
1st Migration 2nd Migration
1. From “here”:
November 1, 2011 6
• mobile FDD• Ethernet backhaul• that “last E1”
From Legacy Sync ...to SyncE
RNCN x E1
“all SyncE”
Network
RNC
E1
N x E1
Operator Network
(TDM)
Ethernet
Network
SyncE flow
From Legacy Sync ...to PTP
RNCN x E1
Network
(3rd party)
PTP
masterPTP
slave
PTP timing flow
(G.8265.1)
RNC
E1
N x E1
Operator Network
(TDM)
Network
(3rd party)
From Legacy Sync ...to Hybrid networks
RNCN x E1
Network
(3rd party)
PTP
masterPTP
slave
RNC
E1
N x E1
Operator Network
(TDM)
Network
(3rd party)
SyncE
PTP timing flow
SyncE flow
2. Timely Progress
November 1, 2011 10
• Network limits G.8261.1• Packet Equipment Clock model G.8263• PEC “knowledge base” G.8260
Output limits for a packet Clock(frequency / FDD applications)
112G / 3G / LTE mobile requires 50ppb frequency stability
2us
18us
MTIE
2us +16ppb
18us + 50ppb
Or...
18us + 16ppb
50ppb
9 µs
Observation
interval τ (s)32
MTIE
0.2
16ppb
1125
18 µs
640.05
G.8261.1 Sneak Preview
Factors impacting Packet clock “performance”:
•Local oscillator stability
•Accuracy of timestamps
•Sync rate
•PDV characteristics of the network
•PDV filtering in the packet clock
•Packet selection algorithm
•Number of synchronization flows in Ensemble
•Ensemble combining algorithm
•Adaptive clock recovery algorithm
None Some of above being standardized
2010
2011
Model of a frequency PEC (G.8263)
November 1, 2011 13
Noise
Amplitude
(us)P
acket s
ele
ctio
n
PECoutput
Ensem
ble
2nd order PLL
Master_1
Master_n
...
PEC output
for worst case input
November 1, 2011 14
Worst case
Noise
Amplitude
OutputFFO2nd order
PLL
Packet Equipment Clock model
– not just ”secret plans and clever tricks”
November 1, 2011 15
0
10
20
30
40
50
60
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Max o
utp
ut
FF
O f
or
all f
req
uen
cie
s (
pp
b)
Time Constant (seconds)
100µs PDV Triangular noise
200µs PDV Triangular noise
500µs PDV Triangular noise
16ppb objective
5000s line
PEC Time constant guarentees a
noise tolerance
10/25/10 16
3. Supporting PTP migration
(Frequency FDD)
• PDV metrics
• Clock output monitoring
• Live network deployments
Monitoring of the Synchronization
network
Mobile networks:
– need synchronization <50ppb
– May be >10,000 nodes
– Some must migrate to IEEE1588
PECs viewed with suspicion
– Are fairly new
– Telecom standards still evolving
– Performance partially dependent on network conditions
17Operators need to be conservative
Packet Metrics – work in progress ...
10/25/10 18
0
10
20
30
40
50
60
1 10 100 1000 10000
Available, but...
Hard to interpret...
Needs mapped to FFO
10/25/10 19
True PEC output monitoring
(Built-in wander meter)
Needs a reference, but...
Fortunately legacy provides
the reference (temporarily).qPEC qSD
H
IEEE1588 Clock Monitoring
Phase
comparitor
Node
Clock
qSD
HqPEC...
TIE
+
MTIE > Mask Alarm
TIE measurement
TIE measurements
21
MTIE measurements
(up to 10 x 24h)
Which of my 10,000 nodes should I monitor?
March 14, 2011 - DRAFT 22
Filter for 1588 ”Mask Crossed”
alarms
Change mask for senarios
Why not monitor all 10,000?
Network management
Supports 1588 clock monitoring
March 14, 2011 - DRAFT 23
Select reference (for Migration use ”last E1”).
Configure MTIE mask and monitoring period
Automate configuration using tools; rules and macros
Set Fall Back List priorities (determines node clock)
E.g. Initially E1 then switch to PTP
Large scale evaluation of IEEE1588
performance (TDM to IEEE1588 migration)
Work flow:
1. Configure clock monitoring (Node clock = E1)
2. Periodically during migration
– filter faults to show synchronization “trouble spots”
– launch clock monitoring on selected node
– optionally “tighten MTIE mask” for scenarios
3. Preference PTP over E1 (Node clock = PTP)
(clock selection process does the “switchover)
4. Decommission the “last E1” from nodes
March 14, 2011 - DRAFT 24Perhaps leave a few measurements points...
10/25/10 25
4. Future Directions(phase synchronization TDD/LTE adv.)
•Phase requirements (new network)
•Telecoms boundary clock
•Network phase & service phase
10/25/10 26
Telecom Phase and frequency solutions are very different...
... but same management (SSM) for both
Prepare for the ”phase migration”
LTE-TDD and LTE-advanced
– Require phase synchronization
– 1us or better accuracy
GPS extension using IEEE1588
– Standardization work ongoing
– Nodes ”T-BC Hardware Ready”
10/25/10 27
RNC
Network(Ethernet)GPS
GPS
PPS
Initially islands around GPS
SyncE
Microwave
links
Synchronization TopologyProvisioned, uses ”SSM” for ring protection
Migration to support phase means
– Upgrading PRC to PRTC
– Upgrading EEC’s to T-BC’s
10/25/10 28
SDHSyncE
10GRNC
L3L2L3 L1
Topology management is the same for phase
PR CT
Network timing flow
Network phase flow
Phase sync needs 4 timestamps
Delivery mechanisms differ...
November 1, 2011 29
Master Slave
T1T2T3
T4
T4
IEEE1588 Delay
Request Mechanism
T2T1
T4T3
ESMC timestamp mechanism
(Phase Synchronous Ethernet)
a)
b)
c)
T2T1
T4T3 IEEE1588 Pdelay Request Mechanism
(Sync message free for ”Service Phase”)
10/25/10 30
L1 SyncE
L2 Ethernet
T-BC
L3 IPService
Phase(s)
Network
Phase
Network
Frequency
Telecoms BC
Service GW
Telecoms BC
Service GW
SGW SGW
Service phase and Network Phase
E.g. When a carrier has several mobile customers...
“SSM” and the sync hierarchy remain the same!10/25/10 31
5. Conclusions
Everyone is leaving “Legacy Land” on...
– SyncE; PTP; Hybrid (or anything that floats)
Packet Equipment clocks - ready for Ethernet prime-time
PEC output monitoring gives a confidence boost
– Last service of the “Last E1”
– Large-scale, live network, field monitoring is available
Prepare for 2nd migration
– Phase support needs new clocks (Nodes “HW phase ready”)
Thank you!
Questions?
Enabling the mobile Internet
November 1, 2011 33
10/25/10 34
A PEC’s got to know it’s limitations
PTSF