The active using virtualization technologies which are applied in the construction of network infrastructure provides a possibility of applying new methods and tools for managing the components of such infrastructures. The paper discusses the use of such tools as the Ansible platform, the NCCLIENT and PyEZ libraries for the dynamic management of virtual components of a hardware-software workbench as part of the laboratory complex of softwaredefined network infrastructures of RTF Ural Federal University (UrFU).
Ansible) [
- special libraries for universal programming languages, such as, for example, the NETCONF Client (NCCLIENT)
[
In following sections, we will discuss the use of tools such as the Ansible platform, the NCCLIENT and PyEZ libraries for dynamically managing the virtual components of the hardware-software laboratory bench of the software-defined network infrastructures of UrFU. 2
UrFU's SDN workbench Infrastructure is based on hardware complex, which was previously used for laboratory VoIP
classes providing. It includes both virtual and hardware components of the network infrastructure, using of which, as was
shown earlier [
To organize serial connections between the routers of the laboratory complex, fractional digital streams are used, formed
in the E1 channel, the parameters of which are defined in ITU-T G.704 [
The complex described above was used for laboratory work as part of the testing of the optional course “Fundamentals of IP-telephony”. During the performance of these works, 2-4 teams of students worked simultaneously, whose task was to configure virtual PBXs and other infrastructure components in order to ensure incoming and outgoing calls using standard (Cisco Unified IP Phone 7912G / 7941G / 7942G) or video ( Cisco Unified IP Phone 9971) Phones.
Performed tests showed the full performance of the laboratory complex and confirmed the correctness of the solutions
that were chosen for its construction. To assess the feasibility of implementing a system that could provide dynamic control
of the virtual channel capacity during an audio-video call, it was necessary to determine the characteristics of the
information stream devoted to provide it. The measurements were carried out in one of the virtual layers of the topology
shown in Fig. 1 using Cisco Unified IP Phone 9971 phones, in which the H.264 / AVC codec is used for encoding /
decoding an audio-video stream [
The measurements results are shown in Fig. 2.a and 2.b of the information flow characteristics during the transmission of quasi-static and dynamic VoIP images, respectively. The presented histograms were generated according to the results of 500 measurements with three values of the throughput of the virtual channel that connects vR1 and vR2: 128, 192 and 256 Kbps (vertical hatching, inclined hatching and uniform filling, respectively).
The obtained results made it possible to establish that during a quasi-static images exposing for VoIP phones, they formed digital stream, by average data rate of (110 ± 10) Kbps. To transmit such digital stream 128 Kbps throughput channel was sufficient.
It was also found that during a dynamic images exposing for VoIP phones, they formed digital stream by average data rate of (170 ± 10) Kbps. Such stream was transmitted without loss through channels with a bandwidth of 192 and 256 Kbps. Examples has shown in Fig. 3.a-3.c of such call's by using a virtual channels with a bandwidth of 64, 128 and 192 Kbps, respectively.
During those tests, it was found that the transmission led to losses and deterioration of image quality due to overloading of this channel of the information flow of an audio-video call with a moving image through a channel with a bandwidth of 128 Kbps. It should also be noted that a channel provides almost error-free transmission of such a stream with a bandwidth of 192 Kbps. The results, obtained by those tests were used to develop a program that is able to control the bandwidth of a virtual channel by redistributing the resource of the corresponding component of the underlay network (in this case, the physical channel) between the virtual layer's components. 3
Among configuration management platforms such as Ansible, Saltstack, Puppet, or Chef, Ansible has been the most
popular over the past 5 years [
In the event of problems associated with a managed device, and not with syntax or other errors in the playbook, Ansible
monitors errors separately for each device. If an error occurs with a specific device, Ansible stops processing tasks for that
device. However, if there are no errors on other devices, the tasks for these devices will be completed [
To manage the constructed UrFU's SDN workbench, playbook and template files were developed, which allow to change the bandwidth of the communication channel of two virtual IP PBX.
As a result, changing the bandwidth of the communication channel of two virtual routers is reduced to the execution of one command: ansible-playbook slot-settings.yaml -e 'slot = 1-4', where the -e parameter passes the value of the slot variable, which determines the number of time-slots used.
Automation platforms, considered on the example of Ansible, are convenient for efficient configuration of many devices and are used, as a rule, in cases when the required state of the system after configuration is determined in advance. To solve the problem posed in this article, it is important that the system itself can respond to changes and move from one state to another depending on certain factors. Therefore, the use of these tools is impractical in solving this problem. 4
The NETCONF and RESTCONF protocols are actively used today to automate the management of network equipment
[25]. The use of these protocols allows us to translate the solution of network equipment control problems to a qualitatively
new level, making possible dynamic reconfiguration of the network infrastructure in real time [
Initially, the scripts were prepared using the Python library - NCCLIENT, working on the protocol NETCONF - a network protocol for managing devices. Using the library simplifies equipment setup and management. The library is exported to a Python script and connection parameters via SSH are set to connect to devices.
The developed technological scripts allow reading, configuring and monitoring routers without using the command line. The use of these scripts greatly simplified the equipment configuration processes, and made it possible to get operational access to such parameters of the virtual port of the router as the current value of the transmission speed on this port and the number of packets that were dropped due to an overload.
The connect function is used, in the parameters of which the necessary data is connected to establish a connection with the device. The current device configuration is written to the xml Dom using the get_config function. The full list is given in the repository of the prepared technological scripts [27]. Another tool is the PyEZ library [28] for automating the management of network equipment. The PyEZ library is similar to NCCLIENT one except it supports only Juniper devices. The narrow focus of this library allows you to form a more compact program code.
In the process of comparing the tools the development of a technological script is the most suitable in Python using libraries based on the NETCONF Protocol to solve the problem 'of dynamic management of virtual components.
The next stage of the work was the preparation of the controller program which allows to dynamically adapt the bandwidth of the virtual channel to the requirements of the transmitted signal. 5
The controller program is a Python script that uses the procedures of the PyEZ library described above. The script runs on the workstation which is located in the equipment control network.
The controller operation algorithm includes the following steps: channel status monitoring; compute of parameters; accepting decision on the need to change the channel capacity; bandwidth change if necessary.
At the monitoring stage, using the methods of the PyEZ library and the NETCONF protocol, the controller requests the following parameters from the router: serial port speed, interface transfer rate and number of lost packets. The measured data rate values are recorded in two parameters: one contains the current speed value, the other contains the maximum transmission rate since the channel bandwidth changes. The values of the lost packets are written to a buffer, which is organized as a queue with a fixed length. Thus, at each moment of time, the average value is calculated of discarded packets for a certain period of time. As a result of monitoring and calculation of the parameters, the following values are determined: - Average number of dropped packets; - Difference between maximum and current data rates.
At the next stage, a decision is made on the need for bandwidth changes. The controller increases the transmission channel if the average number of dropped packets exceeds the value X. In the event that the difference between the maximum and current data transfer rate exceeds the value Y, the controller reduces the transmission channel. The block diagram is shown in Figure 4 of the controller program algorithm.
The virtual channel bandwidth is set to 64 Kbps and a static image is used in the initial phase of the test. Since this bandwidth is not enough to transmit a static image, the Drop Rate exceeds the threshold value and the controller decides to increase the information speed limit to 128 Kbps (point " A” in the diagram). Then the data transmission process becomes stationary 110 Kbps. The presence of such fluctuations is explained by the fact that the tests were carried out for a sufficiently long time interval (from 3 to 10 minutes) and video telephones recorded the movements of test participants in the classroom. Approximately one minute after the start of the test, a moving object (a floor fan) was placed in the field of the video phone cameras, which, as shown above, increased the average information speed to 170 Kbps. In this case, as in the previous one, the Drop Rate exceeds the threshold value and the controller decides to increase the maximum information speed, this time to the value of 192 Kbps (point “B” on the diagram). After another minute, the moving object was removed from the field of video phone cameras, and the average value decreased again to 110 Kbps of the information speed. In this case, the discrepancy between the limiting and current speeds exceeds the threshold value and the controller decides to reduce the limiting information speed to 128 Kbps (point “C” in the diagram).
Multiple repeated tests showed similar results, which confirm the performance of hardware and software solutions that were the basis for the described workbench. The process did not require additional resources for the developing a hardware and software workbench alone; its programming also did not cause much difficulty due to the presence of good documentation of the NCCLIENT and PyEZ libraries and a large number of examples of their use. All this made it possible to carry out work on the creation and commissioning of the workbench in a fairly short time (May-October 2019). 6
The active use of virtualization technologies which are applied in the construction of modern network infrastructures makes it necessary to use new methods for managing the components of these structures. The introduction of these methods allows you to automate the component management process, which helps to increase the efficiency of using the network infrastructure's resources and make it possible dynamic changing infrastructure's configuration. Today, these features can be realized through the use of software platforms designed for configuration management or applications based on special libraries that are designed for many of the popular programming languages. The paper discusses the experience of using such tools as the Ansible platform, the Python libraries of the NCCLIENT and PyEZ for the dynamic management of virtual components of the laboratory workbench in the complex of software-configured network infrastructures of RTF UrFU. Due to the availability and maturity of these tools, the work was fully completed in a short time on creating and programming the laboratory workbench. The experience gained during the of these works fulfilling allows us to conclude that the future use of such tools is promising and fruitful for automation of network equipment management. Acknowledgements This research paper is supported by 211 acts of the Government of the Russian Federation, agreement No. 02.A03.21.0006. 24. IETF RFC8040 RESTCONF Protocol https://tools.ietf.org/html/rfc8040 25. Wallin, S., & Wikström, C. Automating network and service configuration using NETCONF and YANG”. In LISA’11 Proceedings of the 25th International Conference on Large Installation System Administration (2011)(p. 22). 26. Netconf Central http://www.netconfcentral.org/ 27. RTF PKS repository https://github.com/KlimovaAlina/SDN 28. Junos PyEZ Documentation (https://www.juniper.net/documentation/product/en_US/junos-pyez)