`gnpy` is an open-source, community-developed library for building route planning and optimization tools in real-world mesh optical networks.
gnpy is:
- a sponsored project of the OOPT/PSE working group of the Telecom Infra Project
- fully community-driven, fully open source library
- driven by a consortium of operators, vendors, and academic researchers
- intended for rapid development of production-grade route planning tools
- easily extensible to include custom network elements
- performant to the scale of real-world mesh optical networks
Documentation: https://gnpy.readthedocs.io
There are weekly calls about our progress. Newcomers, users and telecom operators are especially welcome there. We encourage all interested people outside the TIP to join the project.
- all releases are available via GitHub
- the master branch contains stable, validated code. It is updated from develop on a release schedule determined by the OOPT-PSE Working Group.
- the develop branch contains the latest code under active development, which may not be fully validated and tested.
Our Docker images contain everything needed to run all examples from this guide. Docker transparently fetches the image over the network upon first use. On Linux and Mac, run:
$ docker run -it --rm --volume $(pwd):/shared telecominfraproject/oopt-gnpy
root@bea050f186f7:/shared/examples#
On Windows, launch from Powershell as:
PS C:\> docker run -it --rm --volume ${PWD}:/shared telecominfraproject/oopt-gnpy
root@89784e577d44:/shared/examples#In both cases, a directory named examples/ will appear in your current working directory.
GNPy automaticallly populates it with example files from the current release.
Remove that directory if you want to start from scratch.
Note: gnpy supports Python 3 only. Python 2 is not supported. gnpy requires Python ≥3.6
Note: the gnpy maintainers strongly recommend the use of Anaconda for managing dependencies.
It is recommended that you use a "virtual environment" when installing gnpy. Do not install gnpy on your system Python.
We recommend the use of the Anaconda Python distribution which comes with many scientific computing
dependencies pre-installed. Anaconda creates a base "virtual environment" for
you automatically. You can also create and manage your conda "virtual
environments" yourself (see:
https://conda.io/docs/user-guide/tasks/manage-environments.html)
To activate your Anaconda virtual environment, you may need to do the following:
$ source /path/to/anaconda/bin/activate # activate Anaconda base environment
(base) $ # note the change to the promptYou can check which Anaconda environment you are using with:
(base) $ conda env list # list all environments
# conda environments:
#
base * /src/install/anaconda3
(base) $ echo $CONDA_DEFAULT_ENV # show default environment
baseYou can check your version of Python with the following. If you are using Anaconda's Python 3, you should see similar output as below. Your results may be slightly different depending on your Anaconda installation path and the exact version of Python you are using.
$ which python # check which Python executable is used
/path/to/anaconda/bin/python
$ python -V # check your Python version
Python 3.6.5 :: Anaconda, Inc.From within your Anaconda Python 3 environment, you can clone the master branch of the gnpy repo and install it with:
$ git clone https://github.com/Telecominfraproject/oopt-gnpy # clone the repo
$ cd oopt-gnpy
$ python setup.py install # installTo test that gnpy was successfully installed, you can run this command. If it
executes without a ModuleNotFoundError, you have successfully installed
gnpy.
$ python -c 'import gnpy' # attempt to import gnpy
$ pytest # run testsgnpy is a library for building route planning and optimization tools.
It ships with a number of example programs. Release versions will ship with fully-functional programs.
Note: If you are a network operator or involved in route planning and optimization for your organization, please contact project maintainer Jan Kundrát <jan.kundrat@telecominfraproject.com>. gnpy is looking for users with specific, delineated use cases to drive requirements for future development.
This example demonstrates how GNPy can be used to check the expected SNR at the end of the line by varying the channel input power:
By default, this script operates on a single span network defined in examples/edfa_example_network.json
You can specify a different network at the command line as follows. For example, to use the CORONET Global network defined in examples/CORONET_Global_Topology.json:
$ ./examples/transmission_main_example.py examples/CORONET_Global_Topology.json
It is also possible to use an Excel file input (for example examples/CORONET_Global_Topology.xls). The Excel file will be processed into a JSON file with the same prefix. For further instructions on how to prepare the Excel input file, see Excel_userguide.rst.
The main transmission example will calculate the average signal OSNR and SNR
across network elements (transceiver, ROADMs, fibers, and amplifiers)
between two transceivers selected by the user. Additional details are provided by doing transmission_main_example.py -h. (By default, for the CORONET Global
network, it will show the transmission of spectral information between Abilene and Albany)
This script calculates the average signal OSNR = Pch/Pase and SNR = Pch/(Pnli+Pase).
Pase is the amplified spontaneous emission noise, and Pnli the non-linear interference noise.
Design and transmission parameters are defined in a dedicated json file. By default, this information is read from examples/eqpt_config.json. This file defines the equipment libraries that can be customized (EDFAs, fibers, and transceivers).
It also defines the simulation parameters (spans, ROADMs, and the spectral information to transmit.)
The EDFA equipment library is a list of supported amplifiers. New amplifiers can be added and existing ones removed. Three different noise models are available:
'type_def': 'variable_gain'is a simplified model simulating a 2-coil EDFA with internal, input and output VOAs. The NF vs gain response is calculated accordingly based on the input parameters:nf_min,nf_max, andgain_flatmax. It is not a simple interpolation but a 2-stage NF calculation.'type_def': 'fixed_gain'is a fixed gain model. NF == Cte == nf0 if gain_min < gain < gain_flatmax'type_def': Noneis an advanced model. A detailed JSON configuration file is required (by default examples/std_medium_gain_advanced_config.json). It uses a 3rd order polynomial where NF = f(gain), NF_ripple = f(frequency), gain_ripple = f(frequency), N-array dgt = f(frequency). Compared to the previous models, NF ripple and gain ripple are modelled.
For all amplifier models:
| field | type | description |
|---|---|---|
type_variety |
(string) | a unique name to ID the amplifier in the JSON/Excel template topology input file |
out_voa_auto |
(boolean) | auto_design feature to optimize the amplifier output VOA. If true, output VOA is present and will be used to push amplifier gain to its maximum, within EOL power margins. |
allowed_for_design |
(boolean) | If false, the amplifier will not be picked by auto-design but it can still be used as a manual input (from JSON or Excel template topology files.) |
The fiber library currently describes SSMF and NZDF but additional fiber types can be entered by the user following the same model:
| field | type | description |
|---|---|---|
type_variety |
(string) | a unique name to ID the fiber in the JSON or Excel template topology input file |
dispersion |
(number) | (s.m-1.m-1) |
gamma |
(number) | 2pi.n2/(lambda*Aeff) (w-2.m-1) |
The transceiver equipment library is a list of supported transceivers. New transceivers can be added and existing ones removed at will by the user. It is used to determine the service list path feasibility when running the path_request_run.py routine.
| field | type | description |
|---|---|---|
type_variety |
(string) | A unique name to ID the transceiver in the JSON or Excel template topology input file |
frequency |
(number) | Min/max as below. |
mode |
(number) | A list of modes supported by the transponder. New modes can be added at will by the user. The modes are specific to each transponder type_variety. Each mode is described as below. |
The modes are defined as follows:
| field | type | description |
|---|---|---|
format |
(string) | a unique name to ID the mode |
baud_rate |
(number) | in Hz |
OSNR |
(number) | min required OSNR in 0.1nm (dB) |
bit_rate |
(number) | in bit/s |
roll_off |
(number) | Not used. |
tx_osnr |
(number) | In dB. OSNR out from transponder. |
cost |
(number) | Arbitrary unit |
Simulation parameters are defined as follows.
Auto-design automatically creates EDFA amplifier network elements when they are
missing, after a fiber, or between a ROADM and a fiber. This auto-design
functionality can be manually and locally deactivated by introducing a Fused
network element after a Fiber or a Roadm that doesn't need amplification.
The amplifier is chosen in the EDFA list of the equipment library based on
gain, power, and NF criteria. Only the EDFA that are marked
'allowed_for_design': true are considered.
For amplifiers defined in the topology JSON input but whose gain = 0
(placeholder), auto-design will set its gain automatically: see power_mode in
the Spans library to find out how the gain is calculated.
Span configuration is performed as follows. It is not a list (which may change in later releases) and the user can only modify the value of existing parameters:
| field | type | description |
|---|---|---|
power_mode |
(boolean) | If false, gain mode. Auto-design sets
amplifier gain = preceding span loss,
unless the amplifier exists and its
gain > 0 in the topology input JSON.
If true, power mode (recommended for
auto-design and power sweep.)
Auto-design sets amplifier power
according to delta_power_range. If the
amplifier exists with gain > 0 in the
topology JSON input, then its gain is
translated into a power target/channel.
Moreover, when performing a power sweep
(see power_range_db in the SI
configuration library) the power sweep
is performed w/r/t this power target,
regardless of preceding amplifiers
power saturation/limitations. |
delta_power_range_db |
(number) | Auto-design only, power-mode only. Specifies the [min, max, step] power excursion/span. It is a relative power excursion w/r/t the power_dbm + power_range_db (power sweep if applicable) defined in the SI configuration library. This relative power excursion is = 1/3 of the span loss difference with the reference 20 dB span. The 1/3 slope is derived from the GN model equations. For example, a 23 dB span loss will be set to 1 dB more power than a 20 dB span loss. The 20 dB reference spans will always be set to power = power_dbm + power_range_db. To configure the same power in all spans, use [0, 0, 0]. All spans will be set to power = power_dbm + power_range_db. To configure the same power in all spans and 3 dB more power just for the longest spans: [0, 3, 3]. The longest spans are set to power = power_dbm + power_range_db + 3. To configure a 4 dB power range across all spans in 0.5 dB steps: [-2, 2, 0.5]. A 17 dB span is set to power = power_dbm + power_range_db - 1, a 20 dB span to power = power_dbm + power_range_db and a 23 dB span to power = power_dbm + power_range_db + 1 |
max_fiber_lineic_loss_for_raman |
(number) | Maximum linear fiber loss for Raman amplification use. |
max_length |
(number) | Split fiber lengths > max_length. Interest to support high level topologies that do not specify in line amplification sites. For example the CORONET_Global_Topology.xls defines links > 1000km between 2 sites: it couldn't be simulated if these links were not split in shorter span lengths. |
length_unit |
"m"/"km" | Unit for max_length. |
max_loss |
(number) | Not used in the current code implementation. |
padding |
(number) | In dB. Min span loss before putting an attenuator before fiber. Attenuator value Fiber.att_in = max(0, padding - span_loss). Padding can be set manually to reach a higher padding value for a given fiber by filling in the Fiber/params/att_in field in the topology json input [1] but if span_loss = length * loss_coef + att_in + con_in + con_out < padding, the specified att_in value will be completed to have span_loss = padding. Therefore it is not possible to set span_loss < padding. |
EOL |
(number) | All fiber span loss ageing. The value
is added to the con_out (fiber output
connector). So the design and the path
feasibility are performed with
span_loss + EOL. EOL cannot be set
manually for a given fiber span
(workaround is to specify higher
con_out loss for this fiber). |
con_in,
con_out |
(number) | Default values if Fiber/params/con_in/out is None in the topology input description. This default value is ignored if a Fiber/params/con_in/out value is input in the topology for a given Fiber. |
{
"uid": "fiber (A1->A2)",
"type": "Fiber",
"type_variety": "SSMF",
"params":
{
"type_variety": "SSMF",
"length": 120.0,
"loss_coef": 0.2,
"length_units": "km",
"att_in": 0,
"con_in": 0,
"con_out": 0
}
}ROADMs can be configured as follows. The user can only modify the value of existing parameters:
| field | type | description |
|---|---|---|
target_pch_out_db |
(number) | Auto-design sets the ROADM egress channel
power. This reflects typical control loop
algorithms that adjust ROADM losses to
equalize channels (eg coming from different
ingress direction or add ports)
This is the default value
Roadm/params/target_pch_out_db if no value
is given in the Roadm element in the
topology input description.
This default value is ignored if a
params/target_pch_out_db value is input in
the topology for a given ROADM. |
add_drop_osnr |
(number) | OSNR contribution from the add/drop ports |
restrictions |
|
If non-empty, keys If no booster should be placed on a degree,
insert a |
The SpectralInformation object can be configured as follows. The user can
only modify the value of existing parameters. It defines a spectrum of N
identical carriers. While the code libraries allow for different carriers and
power levels, the current user parametrization only allows one carrier type and
one power/channel definition.
| field | type | description |
|---|---|---|
f_min,
f_max |
(number) | In Hz. Carrier min max excursion. |
baud_rate |
(number) | In Hz. Simulated baud rate. |
spacing |
(number) | In Hz. Carrier spacing. |
roll_off |
(number) | Not used. |
tx_osnr |
(number) | In dB. OSNR out from transponder. |
power_dbm |
(number) | Reference channel power. In gain mode (see spans/power_mode = false), all gain settings are offset w/r/t this reference power. In power mode, it is the reference power for Spans/delta_power_range_db. For example, if delta_power_range_db = [0,0,0], the same power=power_dbm is launched in every spans. The network design is performed with the power_dbm value: even if a power sweep is defined (see after) the design is not repeated. |
power_range_db |
(number) | Power sweep excursion around power_dbm. It is not the min and max channel power values! The reference power becomes: power_range_db + power_dbm. |
sys_margins |
(number) | In dB. Added margin on min required transceiver OSNR. |
The transmission_main_example.py script propagates a spectrum of channels at 32 Gbaud, 50 GHz spacing and 0 dBm/channel.
Launch power can be overridden by using the --power argument.
Spectrum information is not yet parametrized but can be modified directly in the eqpt_config.json (via the SpectralInformation -SI- structure) to accommodate any baud rate or spacing.
The number of channel is computed based on spacing and f_min, f_max values.
An experimental support for Raman amplification is available:
$ ./examples/transmission_main_example.py \
examples/raman_edfa_example_network.json \
--sim examples/sim_params.json --show-channelsConfiguration of Raman pumps (their frequencies, power and pumping direction) is done via the RamanFiber element in the network topology. General numeric parameters for simulaiton control are provided in the examples/sim_params.json.
Use examples/path_requests_run.py to run multiple optimizations as follows:
$ python path_requests_run.py -h
Usage: path_requests_run.py [-h] [-v] [-o OUTPUT] [network_filename] [service_filename] [eqpt_filename]The network_filename and service_filename can be an XLS or JSON file. The eqpt_filename must be a JSON file.
To see an example of it, run:
$ cd examples
$ python path_requests_run.py meshTopologyExampleV2.xls meshTopologyExampleV2_services.json eqpt_config.json -o output_file.jsonThis program requires a list of connections to be estimated and the equipment
library. The program computes performances for the list of services (accepts
JSON or Excel format) using the same spectrum propagation modules as
transmission_main_example.py. Explanation on the Excel template is provided in
the Excel_userguide.rst. Template for
the JSON format can be found here: service-template.json.
gnpy is looking for additional contributors, especially those with experience
planning and maintaining large-scale, real-world mesh optical networks.
To get involved, please contact Jan Kundrát <jan.kundrat@telecominfraproject.com> or Gert Grammel <ggrammel@juniper.net>.
gnpy contributions are currently limited to members of TIP. Membership is free and open to all.
See the Onboarding Guide for specific details on code contributions.
See AUTHORS.rst for past and present contributors.
Data Centers are built upon interchangeable, highly standardized node and network architectures rather than a sum of isolated solutions. This also translates to optical networking. It leads to a push in enabling multi-vendor optical network by disaggregating HW and SW functions and focusing on interoperability. In this paradigm, the burden of responsibility for ensuring the performance of such disaggregated open optical systems falls on the operators. Consequently, operators and vendors are collaborating in defining control models that can be readily used by off-the-shelf controllers. However, node and network models are only part of the answer. To take reasonable decisions, controllers need to incorporate logic to simulate and assess optical performance. Hence, a vendor-independent optical quality estimator is required. Given its vendor-agnostic nature, such an estimator needs to be driven by a consortium of operators, system and component suppliers.
Founded in February 2016, the Telecom Infra Project (TIP) is an engineering-focused initiative which is operator driven, but features collaboration across operators, suppliers, developers, integrators, and startups with the goal of disaggregating the traditional network deployment approach. The group’s ultimate goal is to help provide better connectivity for communities all over the world as more people come on-line and demand more bandwidth- intensive experiences like video, virtual reality and augmented reality.
Within TIP, the Open Optical Packet Transport (OOPT) project group is chartered with unbundling monolithic packet-optical network technologies in order to unlock innovation and support new, more flexible connectivity paradigms.
The key to unbundling is the ability to accurately plan and predict the performance of optical line systems based on an accurate simulation of optical parameters. Under that OOPT umbrella, the Physical Simulation Environment (PSE) working group set out to disrupt the planning landscape by providing an open source simulation model which can be used freely across multiple vendor implementations.
We believe that openly sharing ideas, specifications, and other intellectual property is the key to maximizing innovation and reducing complexity
TIP OOPT/PSE's goal is to build an end-to-end simulation environment which defines the network models of the optical device transfer functions and their parameters. This environment will provide validation of the optical performance requirements for the TIP OLS building blocks.
- The model may be approximate or complete depending on the network complexity. Each model shall be validated against the proposed network scenario.
- The environment must be able to process network models from multiple vendors, and also allow users to pick any implementation in an open source framework.
- The PSE will influence and benefit from the innovation of the DTC, API, and OLS working groups.
- The PSE represents a step along the journey towards multi-layer optimization.
gnpy is distributed under a standard BSD 3-Clause License.
See LICENSE for more details.