Mentions légales du service

Skip to content
Snippets Groups Projects
Select Git revision
  • dev/all_in_one
  • master default protected
  • dev/docker-env
  • fix/flent-fping
  • dev/restore-network-back-pressure
5 results
Name Last commit Last update
apps
inputs
notebooks
tansiv
.dockerignore
.gitignore
LICENSE.txt
MANIFEST.in
README.rst
setup.cfg
setup.py
README.rst

python-tansiv

Python project serving as an entry point for a collection of helper functions to evaluate a Tansiv System (TS).

What's a Tansiv System (TS)

At the infrastructure level, a TS is composed of some QEMU virtual machines (VMs). But, instead of communicating through a regular network those VMs communicate through Simgrid. A TS system is thus an hybrid system where your real application runs on top of a simulated network.

  • a Tansiv System:

    +------------+                       +------------+
|            |                       |            |
|   my app   |                       |   my app   |
|   process1 | <==== SIMGRID =====>  |   process2 |
|            |                       |            |
|        vm1 |                       |         vm2|
+------------+                       +------------+

  • Sometime for comparison study we'll consider a No Tansiv System (NTS). In this case the virtual machines are linked using a regular linux bridge (in the future we may want to evaluate NTS with other options offered by QEMU like sockets)

    +------------+                       +------------+
|            |                       |            |
|   my app   |                       |   my app   |
|   process1 | <== Linux Bridge ==>  |   process2 |
|            |                       |            |
|        vm1 |                       |         vm2|
+------------+                       +------------+

In a docker

laptop) docker run  --device /dev/net/tun --cap-add NET_ADMIN -v $(pwd)/packer:/srv/packer -ti tansiv-ops:latest

NTS

docker) ./notansiv.py --qemu_cmd qemu-system-x86_64 --qemu_mem 1g --qemu_image packer/packer-debian-10-x86_64-qemu/debian-10-x86_64.qcow2 --autoconfig_net

TS

So, What's here ?

A minimal command line is provided, get the help with tang --help.

This CLI basically allows for

  • deploying a TS on Grid'5000 (tan deploy --mode tantap ...)
    • launch some evaluation tasks on top of it (see tan --help for available options)
  • starting a NTS on Grid'5000 (tan deploy --mode tap ...)
  • other util stuffs

How can I install it ?

For instance on G5K (with automatic reservation of one node)

  • One time setup:

    $frontend$) git clone git+https://gitlab.inria.fr/tansiv/python-tansiv.git
$frontend$) cd python-tansiv
$frontend$) virtualenv -p python3 venv
$frontend$) source venv/bin/activate
$frontend$)(venv) pip install -U -e .

  • Launch a TS:

    $frontend$)(venv) tan --env myxp deploy --qemu_image /grid5000/virt-images/debian10-x64-min-2021041419.qcow2 g5k ts inputs/star.xml inputs/deployment_2.xml
# --env myxp             : the directory where some of the experiment data will be saved
# inputs/star.xml: a platform file (simulated machines and network characteristics)
# inputs/deployment_2.xml  : the mapping between virtual machines and host in the plaform file
# image                  : the virtual machine image (will be copied from the local machine to the g5k node)

  • Launch a NTS:

    $frontend$)(venv) tan --env myxp2 deploy g5k nts 2

# The last argument is the number of VMs to launch

  • Benchmark using flent

    $frontend$)(venv) tan --env myxp flent

Analyse (on g5k)

  • (One time setup) Register the python-tansiv virtual environment as a usable jupyter lab kernel.
$frontend$) source venv/bin/activate
$frontend$)(venv) ipython kernel install --name tansiv_venv --user