new optical lab facility @garr: set-up and first results · white rabbit: python+influxdb+grafana...
TRANSCRIPT
New optical Lab facility @GARR: set-up and first results
GLORIA VUAGNIN
INFRASTRUCTURE DEPARTMENT GARR
Prague, September 3rd, 2019
10th CEF Networks workshop
Status of the current infrastructure
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 2
(2011) Huawei 1/10 G• 10G/40Gbps channels not enough for the
needs of the core• Close to the end-of-life• Maintenance (+2Y) ends 2020
(2015) Infinera 10/40/100 G• 500Gbps superchannels• Fine for the capacity needs of Southern
Italy• High power and space consumption
(2017) AW 2x100G among Core PoPs• Buffer solution – it works super fine
(2019) New fiber footprint on North-East Area
• North-Eastern region: closed topology / ARNES direct Interconnection / Lightnetinterconnection
• ECMWF new site in Bologna
MI2MI1
BO1
RM2
BA1
• Open Line System (partially disaggregated) to replace the oldest infrastructure
partially disaggregated == decoupling of photonic elements and transponders
• April 2019: we issued a RFI on open line systems and optical transport network equipment (Juniper, ADVA, ECI, Huawei, Infinera)
• August: tender for DCI (Infinera G30) for an INFN-GARR join project on Distributed Datalake for Science
• By the end of 2019 -> tender (hopefully into GÉANT framework)
How GARR sees its evolution
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 3
We are ready to stand a period of possible inconvenience
(several different DWDM technologies in our network)
in order to build, for the near future, an open infrastructure
Resounding concepts:
Disaggregation
Alien wave
Automation
Analytics
Telemetry
Green
4
What do we have in our mind
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019
We need and we want:
• Acquire knowledge of different available DCI technologies
o Data we are interested: o Reach (in several different scenarios)
o Tolerance to our old infrastructure in the transition time,
o AW working solutions (management and monitoring)
o Figure of merit and acceptance criteria
o New valuable features (analytics …)
• Practice North-Bound Interfaces
o Monitoring & Provisioning
• Start to build an environment for automation
A test facility is a must-have
Building skills and competences for evolution
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 5
An independent-from-production NMS control the lab photonic infrastructure
Optical lab test: the set-up
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 6A
MP
100 km
75 km25 km
25 km
A/D MUX/DEMUX
F-O
AD
M
F-O
AD
M
A/D MUX/DEMUX
Photonic Layer
NODE A3 degree ROADM• Flexgrid• Colorless/Directionless
NODE B3 degree ROADM• Flexgrid• Colorless/Directionless
NODE CF-OADM – 8 channels• Flexgrid• C-Band passthrough
EDFA amplifierExtended C-Band
75 km
DNA LAB
The lab nodes: XTM Infinera
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 7
A
BAmp
C
16 spools of single G.652d fiber
(vendor: Fujikura)
• 8 spools 50km -> 4 pairs
• 8 spools 25 km -> 4 pairs
In total 600km single fiber
The flexible topology changes through a patch panel on the rack top
Fiber spools
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 8
From these PoPs we can configure links that are as long as:
• ~60 km
• ~1000 km
• ~1500 km
Links towards the production environment
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 9
From the lab two cables go down and reach the
RM2 and RM1 PoPs
Paths on the production environment
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 10
End Sites km dB Ila Raman ROADM
RM1 – PI 430 118 4 0 2
PI – FI 136 40 0 2 1
FI – RM2 360 104 3 4 1
Total 926 262 7 6 4
End Sites km dB Ila Raman ROADM
RM1 – RMT 22 10 0 0 2
RMT -- FRA 8 5 0 0 1
FRA – RM2 29 12 0 0 1
Total 59 27 0 0 4
End Sites km dB Ila Raman ROADM
RM1 – PI 430 118 4 0 2
PI–FI-BO1 259 74 0 4 2
BO1-BO–PE 466 144 4 0 3
PE – RM2 361 96 2 0 2
Total 1516 432 10 4 9
Short Metro Ring
Long Ring
Intermediate Ring
ECMWF access to GARR (temporary solution)
For the Italian Data Lake initiative:
Large bandwidth over hundreds of kms
Decisions to make: which transponder for 100GE
11
Site 1 Site 3
Site 2
200Gbps
400Gbps
200Gbps
~700km -220 dB
MI1
BO3
ECMWF
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019
1. Transponder TP100GOTN XTM Infinera
Client: 100G Ethernet
Line: CFP1 100G ACO - QPSK coherent
2. Juniper ACX6360 (and MX240)
Client: 100-Gbps (QSFP28) pluggable interfaces
Line: 200-Gbps CFP2-DCO coherent DWDM pluggable interfaces, which support 100-Gbps QPSK and 200-Gbps 8QAM, and 16QAM modulation options
3. Infinera Groove G30
Client: up to 4x100GbE QSFP28
Line: CHM1 sled – 400G 2xCFP2-ACO (100G QPSK, 150G 8QAM, 200G 16QAM)
Test transponders 100G
12Gloria Vuagnin // CEF 2019// Prague, 09/03/2019
Transponder TP100GOTN XTM Infinera
13
Client: 100G EthernetLine: CFP1 100G ACO - QPSK coherent DWDM
•OTU-FEC-DEG (forward error correction degraded)
•OTU-FEC-EXE (excessive errors, FEC_FAIL from the transponder)
reach
[km]
channel number
(Huawei convention)FEC
adjacent
channels
OSNR
margin
node A
[dB]
OSNR
margin
node B
[dB]
alarms
b2b 43 SD-FEC n 7 7
60 41 SD-FEC n 5 5.5
60 41 SD-FEC y (40) 4.9 5.5
1000 43 SD-FEC n 1 1.5 FEC_EXE
1000 41 SD-FEC y (40) 1.2 1.8
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019
Juniper ACX6360
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 14
Client: 100-Gbps QSFP28Line: 100-Gbps QPSK CFP2-DCO coherent DWDM
•OTU-FEC-DEG (forward error correction degraded)
•OTU-FEC-EXE (excessive errors, FEC_FAIL from the transponder)
•Channel Huawei convention: 43 (193.95), 41 (194.05)
reach
[km]
channel number
(Huawei convention)FEC
adjacent
channels
BER
node A
BER
node B
Qvalue node
A [dB]
Qvalue
node B [dB]alarms
b2b 43 SD-FEC n 4.3E-11 1.1E-11 16.1 16.2
60 41 SD-FEC n 2.5E-07 1.1E-07 13.7 13.9
60 41 SD-FEC y (40) 7.8E-07 1.8E-07 13.6 14.1
1000 43 SD-FEC n 1.00E-03 1.00E-03 7 7 FEC_DEG
1000 41 SD-FEC y (40) 1.60E-03 4.00E-03 9.3 8.4 FEC_EXC
Client: 100GbE QSFP28
Line: CHM1 sled – CFP2-ACO 100G QPSK
Groove test
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 15
ModulationReach
[km]
Channel
number
(Huawei
convention)
adjacent
channels
BER
node A
BER
node B
Qvalue
node A
[dB]
Qvalue
node B
[dB]
OSNR
margin
node A
[dB]
OSNR
margin
node B
[dB]
alarms
60 43 n 1.3E-08 1.2E-08 14.8 15 20 20
60 41 y (40) 2.7E-08 3.5E-08 14.7 14.7 20 20
1000 43 n 3.00E-03 2.20E-03 8.7 9 15.2 14.5
1000 44 n 2.43E-03 5.33E-04 8.9 10.3 15.5 16.5
1000 41 y (40) 6.34E-03 1.58E-03 8 9.5 14.5 15.5
1000 44 y (43) 3.36E-03 1.12E-03 8.6 9.6 15 16.2
1500 43 n 1.50E-02 1.60E-02 6.7 6.5 14 13.6
1500 41 y (40) 5.5 5.6 BER Test errors sync
QPS
K
Client: 100GbE QSFP28
Line: CHM1 sled – CFP2-ACO 150G 8QAM – 200G 16QAM
More modulation tests on Groove
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 16
ModulationReach
[km]
Channel
number
(Huawei
convention)
adjacent
channels
BER
node A
BER
node B
Qvalue
node A
[dB]
Qvalue
node B
[dB]
OSNR
margin
node A
[dB]
OSNR
margin
node B
[dB]
alarms
59 43 n 1.42E-04 2.60E-03 11.1 8.9 20.3 20.7
59 41 y(40) 2.80E-04 3.10E-04 10.7 10.6 20 19.7
926 43 n 1.10E-02 3.40E-03 7 8.6 15.3 15.6 Loss of Frame
926 41 y (40) 5.40E-03 6.40E-03 8.1 7.8 15 14.7 Loss of Frame
59 43 n 4.20E-03 4.20E-03 8.4 8.4 20.5 19.4
59 41 y (40) 8.12E-03 8.12E-03 7.5 7.5 18.8 18.8
926 43 n Loss of Frame
8QA
M16
QA
M
Groove Monitoring using SNMP-influxDB-Grafana
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 17
ECMWF access to GARR (temporary solution)
For the Italian Distributed Data Lake initiative:
Large bandwidth over hundreds of kms
Which transponder for 100GE
18
Site 1 Site 3
Site 2
200Gbps
400Gbps
200Gbps
~700km -220 dB
MI1
BO3
ECMWF
ECMWF IDDLS
INFINERA
XTMOK NOK
Juniper
ACXOK NOK
INIFNERA
GROOVEOK OK
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019
Time & Frequency in Italy
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 19
INRIM in Italy has a dedicated fiber infrastructure
• The aim of the infrastructure is to provide T&F service to Italian Research Institutions (CNR, ASI, INAF) and Industrial Users.
• INRIM main priority is Research
NMI fundamentals: redefinition of the second, better timescale etc.
fundamental physics: geodesy (relativistic or not), VLBI, relativistic experiment, quantum technologies, space.
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 20
WR for a time service: a collaboration with INRIM
• Several research institutions might profit from a source of “INRIMcertified time”
• Not all of them have the same requirements in terms of time accuracy
In July 2019, in collaboration with INRIM, we started to test the White Rabbit protocol over DWDM in our network
• The aim is to check which accuracy we can reach using the WRprotocol on our optical infrastructure (using a pair of fibers, amplifiers and ROADMs)
• We want to measure the average accuracy/stability we can get in different parts of our network
• We want to understand the complexity of operating the WR devices
Playing with a White Rabbit
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 21
LEN1
WRS 1
WRS 2 LEN2
10 MHzIN
10 MHz pps
Free running Clock
Gran Master
Counter
MM M
SS S
Ch1 Ch2Int Ref out
LEN3
Lab set-up
White Rabbit: Python+InfluxDB+Grafana
Gloria Vuagnin // CEF 2019// Prague, 09/03/2019 22
We set-up an optical lab to have the possibility to test new devices and/or solutions in different scenarios.
The lab is a versatile facility, it can work as:
• a completely independent environment
• as part of GARR infrastructure
We built it to enable GARR to evolve its network with:
• a step-by-step approach
• independently checking the value of each solution
Having the possibility to start building our system for:
• a systematic approach to collect data from the network
• good debugging processes
Thank you
Conclusions
23Gloria Vuagnin // CEF 2019// Prague, 09/03/2019