IETF: Align BD generator to new BD model

[mpeuster]

Automatically generate Python class of YANG model using make in this folder. The result of this is vnf_bd.py which will be placed in tng-sdk-benchmark/src/ietf/vnf_bd.py and imported from there.

Generator code is here: https://github.com/mpeuster/tng-sdk-benchmark/blob/feature-31/src/tngsdk/benchmark/ietf/__init__.py

---

Testing in vnf-bench-model:

tng-bench -p peds/upb-vnf-bench-02-01.yml --no-prometheus  --ibbd bds --max-experiments 1 --no-execution

---

• [x] move old BD generator code to src/ietf
• [x] import generated model (try?)
  • [x] ensure that CI (style-check etc. does not break)
  • [x] commit
• [x] populate the new BD model and sync. with Raphael about detected problems

• [ ] check that execution is really optional (run w/o --ibbd flag does not execute new code)

[mpeuster]

Started work in branch feature-31

[mpeuster]

Example of what is generated right now.

The keyed lists are problematic:

---
# PED documentation: https://github.com/sonata-nfv/tng-sdk-benchmark/wiki/PED
#
# This is an example for a profiling experiment descriptor (PED) that
# defines a profiling (benchmarking) experiment for a Suricata VNF
# that is stimulated by traffic traces replayed with TCP replay.
# 
# The tested network service (NSD, VNFDs) can be found in 'nw-services/'.
#
descriptor_version: 0.2  # 5GTANGO: 0.2, SONATA 0.1
vendor: "de.upb"
name: "upb-vnf-bench-02-01"
version: "1.0"
author: "Manuel Peuster, Paderborn University, manuel.peuster@uni-paderborn.de"
description: "NFV Benchmark Experiment"

# Path to the network service we want to profile (relative from PED location)
service_package: "../services/ns-1vnf-ids-suricata"

# Experiment definition
service_experiments:
  - name: "upb_bench_02-01"
    description: "NFV Benchmark Experiment"
    repetitions: 10
    time_limit: 30 # seconds per experiment
    time_warmup: 20
    # NSD to be used (vendor.name.version reference)
    target:
      vendor: "de.upb"
      name: "ns-1vnf-ids-suricata"
      version: "0.1"
    # additional containers for traffic generation/measurements (measurement points)
    measurement_points:
      - name: "mp.output"
        connection_point: "ns:output"
        container: "mpeuster/tng-bench-mp"
        address: "20.0.0.254/24"  # address of the data interface of the MP
      - name: "mp.input"
        connection_point: "ns:input"
        container: "mpeuster/tng-bench-mp"
        address: "20.0.0.1/24"  # address of the data interface of the MP
    # experiment parameters to be tested during the experiment
    experiment_parameters:
      - function: "de.upb.ids-suricata.0.1"
        cmd_start: ["./start.sh small_ruleset",
                    "./start.sh big_ruleset",
                    "./start.sh empty"]
        cmd_stop: "./stop.sh"
        cpu_bw: {"min": 0.1, "max": 1.0, "step": 0.1}  # float: fraction of CPU time in vim-emu
        cpu_cores: ["0"]  # string: this is "cpuset_cpu" in vim-emu e.g. ["0, 1"] -> 2 CPUs
        mem_max: 1024  # int: Memory in MByte
        mem_swap_max: null
        io_bw: null
      - function: "mp.input"
        cmd_start: ["tcpreplay -i data -tK --loop 1000 --preload-pcap /pcaps/smallFlows.pcap",
                    "tcpreplay -i data -tK --loop 1000 --preload-pcap /pcaps/bigFlows.pcap"]
        cmd_stop: "./stop.sh"
        cpu_bw: null  # float: fraction of CPU time in vim-emu
        cpu_cores: "2"  # string: this is "cpuset_cpu" in vim-emu e.g. ["0, 1"] -> 2 CPUs
        mem_max: 1024  # int: Memory in MByte
        mem_swap_max: null
        io_bw: null
      - function: "mp.output"
        cmd_start: null
        cmd_stop: "./stop.sh"  # dump packet counters to logs
        cpu_bw: null  # float: fraction of CPU time in vim-emu
        cpu_cores: "3"  # string: this is "cpuset_cpu" in vim-emu e.g. ["0, 1"] -> 2 CPUs
        mem_max: 1024  # int: Memory in MByte
        mem_swap_max: null
        io_bw: null

[mpeuster]

Issue 1 --> solved! Obsolete.

Use bd_str_json = pybindJSON.dumps(m, mode="ietf") to serialize the model. Important: mode="ietf"

[mpeuster]

Issue 2 --> DONE

In the model, connection points reference the links by their IDs. But the 5GTANGO model, ETSI, and tng-bench, reference the connection point IDs in the links, e.g.,

links': [{
        'id': 'mgmt',
        'connectivity_type': 'E-LAN',
        'connection_points_reference': ['vdu01:mgmt', 'mgmt'],
                ...
    },

Also link id field is unit32 but should be string.

[mpeuster]

Issue 3 --> DONE

Field names in the model. Sometimes plural is used, e.g., scenario/nodes/resources/storage/ units vs. unit (I'd say unit would be correct)

[mpeuster]

Issue 4:

This one only concerns tng-bench.

tng-bench does not support workflow parameters. So the complete command is injected into the implementation field right now. This is not perfect but can be solved later.

Same goes for the prober parameters (for now we us {input: "command", value: "tcpreplay. .."} as a single parameter there).

[mpeuster]

Issue 5 --> solved! Keep as is.

Gym also puts the gym-manager as one part of the scenario. tng-bench is not considered to be part of this, because I consider it outside the scope of the SUT.

This is not a real issue, just a way to handle things in different implementations, I guess.

[mpeuster]

Model updated https://github.com/raphaelvrosa/vnf-bench-model/pull/20