Research and experiment with the tng-bench source code and write ideas on how a development with OSM can take place and which functionality should be added.

Instead of using fixing IP addresses, let's try and experiment with using mDNS addresses (ex, .local)

From now we should monitor the integration process with the benchmarking team.

This issue is designed to track bug fixes and updates done to the tng-bench tool. Major bugs should have their own issues.

This issue is pending on #10
Main task: Update Experiment classes to handle OSM PED file.
Another thing to do is to make sure that the experiment that is generated is exactly the same as OSM VNFD and NSD formats.

Initial Deadline (hopefully not broken): 23.09.2019

To provide better documentation of the benchmarking process using VMs, we should try and preinstall software on Openstack images before they are deployed. This should considerably speed up the experiment times since the images are deployed with the required software, and not installed during the execution of the cloudinit script.

The process of the modification with regards to our scenario should also be well documented.

After testing, --hold is not working in osm pdriver.
This should be fixed.

It was a technical mistake to add the osmclient files under the tngsdk folder.

For correctness and to make pytests run correctly as well, we must move the folder outside the tngsdk folder and put it in the same directory tree level as the tng sdk folder.

Deploy an example docker container (ex. Suricata container) on Openstack.

Responsible person: Bhuvan, with advice from Avi

Handle existing NSD files and add the probe connection points to the NSD.

Example from PED file:

 - function: "mp.input"
        # loop longer than experiment lasts
        cmd_start: ["tcpreplay -i data -tK --loop 1000 --preload-pcap /pcaps/smallFlows.pcap"]
        cmd_stop: "./stop.sh"
        # cpu_bw: null  # float: fraction of CPU time in vim-emu (In OSM no need for this)
        cpu_cores: 1  # 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 (In Openstack no need for this)
        cpu_cores: 1  # 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

This would require that a probe is added to the input of the NSD (i.e. before the first VNF) and a probe to be added to the end of the NSD.

constituent-vnfd:
        -   vnfd-id-ref: hackfest-simplecharm-vnf
            member-vnf-index: '1'
        -   vnfd-id-ref: hackfest-simplecharm-vnf
            member-vnf-index: '2'

DEADLINE: Mon. 14.10.2019 - 08:00

Create the generator that updates the VNFD files to modify params such as CPU and MEM, etc

The OSM generator should also modify the VNFD for the measurement points (probes) according to the parameters provided in the PED file.

The VNFD file is passed already to the generator and is stored in the experiment class. The next task is to pass it to the modifier function to correctly handle the file and modify it.

After writing initial ideas is finished, the driver design document should be written.

It should contain the following (list can be updated):

• Expected technical architecture to be used
• Style of development
• Expected API calls
• Flow diagrams

@team Feel free to add more items if you think I missed something.

Openstack on VM-2 is not working. Issue with OpenVSwitch and q-agent.

Fix is needed ASAP.

Tng-bench is using a tango descriptor generator to modify VNFD and NSD files, we will have to create a new generator to allow for the creation of OSM format files.

This is currently in backlog! Let's focus on working with static VNFD and NSD for now. No one is assigned for this task as of 23.09.2019

Deploy a test network in OSM so that we can know the right configuration.
The network should consist of the following:

• One VNF: Any VNF should do here. Perhaps Suricata as it will be the one we will use for testing our code and driver connection to OSM.
• Two probes: One probe should be on the input side of the VNF, the other should be on the output side. A simple probe here will do more than fine. Any container or VM having TCPreplay or iperf that dumps traffic through the main VNF. The source VNF should be the originator of traffic, and the sink should collect the information that gets passed through.

The purpose of this issue is to be familiar with creating a valid and correct OSM topology that we can then later use to create a correct topology using tng-bench that allows the probes to correctly connect to the main VNF or VNF chain and generate correct benchmarking results.

DEADLINE: 13.09.2019
Please finish the issue before the deadline.

Depends on #10 and #11
Deploy the VNFs after the PED file is processed and the experiment is generated.

Right now just focus on deploying and deleting the experiment. No need to SSH into the VNF images and run commands.

Initial Deadline: 26.09.2019

Make it easy for user to install and run tng-bench from any linux based machine with just one command as a container.

https://github.com/CN-UPB/pg-backflip/blob/master/screenshots/charm-juju-issue.png

Update the tng-bench ped files and create a version that provides the correct parameters to deploy a benchmarking experiment in OSM.

Initial thoughts:
The file should have the following main sections:

• VNFD identifier: the filename or the OSM VNFD identifier. This is the part where the virtual network function is defined.
• NSD Identifier (optional): If left blank, tng-bench should generate this based on the information from the probes section in the ped. The NSD provides information on how the VNF or VNF chain is connected internally, and possibly to the probes.
• Type of experiment.
• Probes: How many probes to connect to the VNF and how.

This list might be updated later on. Updates will be clearly marked.

DEADLINE: 15.09.2019

The OsmClient library that we integrated into our repo uses print statements to output the status of the operations that it is doing. This is not in line with the output standard used by tng-bench. The library should be modified to produce a logging output instead of simple print statements.

Research and experiment with the OSM API connection and write ideas on how a possible connection can be created with tng-bench

Deploy a static VNF/NS using tng-bench. We do not need to create and run any benchmarking tests now, deploying a static VNFD and NSD file and destroying the VNF after tng-bench finishes running.

Now that we have our initial connection ready, we should also start considering unit tests. For this reasons, we should try and see which part of the code can be unit tested.

automate the process of individual user creation on backflip VMs using ansible Playbook

Tests

• user entry in /etc/passwd
• home directory is created for user
• user is able to log in with assigned default password

Tng-bench should try and delete the files it uploaded when it senses a keyboard interrupt before the experiment finishes. This is to reduce the annoyance of having to go to OSM and deleting the VNFD and NSD from there.