Skip to content

Latest commit

 

History

History
226 lines (173 loc) · 6.8 KB

README.md

File metadata and controls

226 lines (173 loc) · 6.8 KB
Raw

svtrace-ansible

svtrace-ansible is a tool used to evaluate the performance of a Seapath standalone infrastructure.

Table of Contents

Introduction

svtrace-ansible focuses on assessing the efficiency and reliability of a Seapath architecture by performing network latency tests. Using IEC61850 protocol, Samples Value are emitted from a publisher machine in destination of a subscriber machine, which will receive, parse, and decode them. The subscriber machine can be a physical machine or a virtual machine hosted on a hypervisor.

NOTE: Each of the machine used are running a Seapath standalone distribution. There is currently no support for a Seapath cluster configuration.

Various latency metrics are computed at various stage of the transit of the Samples Value between each machine, and inside them:

screenshot

  • Network latency: Time elapsed between the emission of an SV (T0) from the publisher machine to the NIC of the subscriber (T3) or of the hypervisor machine (T1).
  • Hypervisor latency: Time elapsed between the reception of an SV on the NIC of the hypervisor machine (T1) to the end of the processing of it (T2). This latency is computed only in the context of a virtualized subscriber.
  • Transit latency: Time elapsed between the end of the processing of an SV on the hypervisor machine (T2) to the reception of the SV on the NIC of the subscriber machine (T3). This latency is computed only in the context of a virtualized subscriber.
  • Subscriber latency: Time elapsed between the reception of an SV on the NIC of the subscriber (T3) machine to the end of the processing of it (T4).

At the end of each test, are generated various graphs and a test report in tests_results directory.

This project is essential for ensuring that the Seapath platform meets the stringent requirements of real-time applications, particularly in the power grid industry.

Features

  • High precision and reliable latency test (+/- 10µs)
  • Identify bottleneck in a Seapath architecture (hypervisor, subscriber)
  • Identify bottleneck inside a Seapath machine

Installation

Requirements

System packages

Make sure your localhost system complies with the following dependencies:

sudo apt install \
docker

Install cqfd:

git clone https://github.com/savoirfairelinux/cqfd.git
cd cqfd
sudo make install

Normally, svtrace-ansible is designed to be run on localhost inside a Docker container. If you wish to run svtrace-ansible directly on the localhost machine, make sure it complies with the following dependencies:

sudo apt install \
ansible \
bash \
python3 \
python3-pip \
rsync

sudo pip install \
matplotlib \
numpy

Seapath machine

A valid IEC61850 PCAP file is needed on the publisher machine. The selection of this file can be achieved using pcap_file inventory variable.

In order to access kernel symbol names, kernel addresses must be exposed with file /proc/sys/kernel/kptr_restrict.

Requirements

Publisher machine

Make sure the following package is installed on the publisher machine:

sudo apt install \
docker \
rsync

The user used by Ansible, targeted by inventory variable ansible_user variable must be configured on the publisher machine to have sudo access.

Hypervisor and subscriber machines

Make sure the following packages are installed on each hypervisor subscriber machines:

sudo apt install \
docker \
psmisc \
rsync \
tshark

The user used by Ansible, targeted by inventory variable ansible_user variable must be configured on the subscriber and hypervisor machines to have sudo access.

Configuration

Time synchronization

Network cards

In order to properly log the Seapath latency, the clock of the network card receiving the Sample Values must be synchronized in PTP. This is done :

  • Automatically if you use the same network card to receive the PTP frames and the SV
  • Manually, using phc2sys to synchronize the SV network card with the PTP one

This is not done by Ansible and must be handled by the user

System

For achieving high accuracy latency measure each machine must be synchronized using the PTP protocol. To implement PTP on Seapath you can follow the detailed procedure in the Seapath Wiki, section "Time synchronization": https://wiki.lfenergy.org/display/SEAP/Time+synchronization

For better PTP performance, we recommend:

  • The isolation, if possible of ptp4l and phc2sys process
  • To use an hardware PTP source
  • To use PTP L2 network transport and P2P delay mechanism
  • To use linreg clock servo for ptp4l and phc2sys configuration
  • In the case of a virtual machine as a subscriber, using kvm_ptp module is highly recommended.

For the svtrace-ansible project, PTP can be configured on Seapath machine without timemaster package. To do so:

  1. Make sure ptp4l and phc2sys utilities are installed on each Seapath machine
  2. timemaster.service is disabled on each Seapath machine:
systemctl disable timemaster.service
  1. PTP variables are correctly configured in your svtrace-ansible inventories files; See README file in inventories/ directory. In particular, make sure ptp_manual_configuration is defined.
  2. Finally, you can run the playbook configure_latency_tests.yaml using:
cqfd -b configure_tests

Synchronizing

General configuration

Seapath machines configuration for the svtrace-ansible project can be set in the dedicated inventories files. See inventories/README for more information.

The configuration can be applied with the configure_latency_tests.yaml playbook:

cqfd -b configure_tests

NOTE: this playbook must be run again each time a variable is modified in a svtrace-ansible inventory.

Usage

After configuration, the tests can be launched with the playbook run_latency_tests.yaml:

cqfd -b run_tests

Test duration can be set achieved using pcap_cycles inventory variable.

The threshold of latency variables can be set in the var file threshold.yaml. To record all SV data, set the threshold value to 0.

For CI purpose, you can also use the launch_tests.sh script.

Release notes

Version v0.1

  • Configuration of the target
  • Hypervisor, subscriber, and total latency measurement
  • Generation of plot results and .adoc report

Version v0.2

  • Add an iteration counter to identify each SV as unique no matter the time of the test
  • Correct a bug where some sample values appeared as duplicated in bpftrace logs
  • Introduce and automate the computation of Tseapath
  • Add Ttot selection in vars/threshold.yaml