Monitoring and Controlling Virtual and Hybrid Test Benches in The Cloud

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Monitoring and Controlling Virtual and Hybrid Test Benches in the

Cloud
A. Lars Stockmann1, B. Jan Wasmund1, C. Sören Reglitz1
1: dSPACE GmbH, Rathenaustraße 26 33102 Paderborn

Abstract: Efficiently operating a test bench is crucial for different test systems of the bench to monitor and control
the success of small- and large-scale projects in the aero- the complete test environment from a single frontend is not
space industry. Traditionally, benches are operated in a lo- possible because multiple proprietary interfaces need to be
cal and often isolated environment, which requires the staff considered.
to be on site. Lockdown situations have made this difficult To improve on this, the Virtual and Hybrid Testing Next
recently, and it became clear that it must also be possible Generation (VHTNG) research project is developing a ge-
to operate the test bench remotely. Making this remote ac- neric and modular architecture for an open test environ-
cess safe, secure, and accessible poses a major challenge ment. The goal is to prepare a set of open standards for the
for suppliers and OEMs alike. A novel approach makes use interfaces required to build this architecture [1]. This in-
of the modularity and openness of test systems following cludes data models and protocol stacks to monitor and con-
the ideas of the Virtual and Hybrid Testing Next Genera- trol such an environment.
tion (VHTNG) research project – a project that aims to cre- Until recently, the VHTNG project has focused primarily
ate a standard for a communication infrastructure by on the local communication in the test bench. However, it
providing standardized transport mechanisms required for has always been ensured that the data models are generic
interoperability. The approach transfers the VHTNG inter- and not limited to a local environment. During the Covid-
faces for unified monitoring, which are currently applied 19 pandemic, it became abundantly clear that having safe
only in local environments, to the cloud. The reasoning be- and efficient remote access to the bench is a must-have for
hind this is that engineers will eventually perform more future test benches. It is a key enabler for flexible testing
tests in the cloud and therefore require location-independ- and efficiently conducting large test campaigns with test
ent test bench services. The approach is based on a dedi- benches allocated on different sites.
cated architecture and a technology stack that enable users
However, providing remote access to monitor and control
to monitor multiple test sites that can consist of one or more
a test bench poses additional challenges, with safety and
test benches via common web browsers. However, the con-
security being the most critical. Adversaries gaining full
trol aspect has not been discussed. Therefore, this paper ex-
access to a test bench could sabotage it to a degree that puts
plores the possibility of using this approach to control test
people’s health at risk. Also, monitoring information, such
benches, too, without sacrificing any of the security-related
as log data, might contain sensitive information. This could
core principles, that is to say, no full access to the test bench
include confidential information about the stage of devel-
and no requirement to expose the test systems directly to
opment, which must not be leaked to competitors in the in-
the Internet. The approach is a direct advancement and is
dustry. Other aspects are maintainability, usability, and, of
based on the same technology stack (i.e., VHTNG and the
course, costs.
Elastic Stack). Just like the previous paper, this work fea-
tures a dedicated requirements analysis and presents the Therefore, VHTNG concepts and models have been ex-
prototype that was used for evaluation. panded to create a secure monitoring solution that can be
deployed in the cloud [2]. This solution makes it possible
Keywords: testing, cloud, remote control to monitor an arbitrary number of test sites, each consisting
of several test benches. In contrast to existing approaches,
it does not impose any additional requirements on an LTM.
1. Introduction This means that it is not required to install additional soft-
ware, nor is there a need to directly connect equipment to
Operating a test bench can be a challenging task. The tester the Internet. This makes the new approach much easier to
must always keep an eye on all the involved laboratory test set up in companies with strict IT restrictions and more se-
means (LTM) and take appropriate actions in case of errors. cure by design. Furthermore, it can leverage the vendor-
In a traditional monolithic test bench, the test system sup- independent nature of VHTNG to include any device if the
plier is responsible for providing the proper tooling for this respective device implements the VHTNG Status Monitor-
task. However, in a heterogeneous test bench, where sys- ing interface. This includes completely virtualized equip-
tems originate from different vendors, a tester must be fa- ment, which itself may be hosted in the cloud.
miliar with all the different ways to operate those systems
using their respective tools. This is not only inconvenient The approach is based on the latest, yet well-established
but can also lead to confusion and requires dedicated train- web technologies and standards. From an end user perspec-
ing for each system. Moreover, unified access to the tive, the approach also supports handheld devices and only
requires a common web browser.
ETTC 2021– European Test & Telemetry Conference
This paper advances the previously presented approach by Requirements 2 to 10 are related to accessing this infor-
adding the possibility to control the test systems. It again mation:
leverages the data models and concepts from VHTNG, REQ2 demands location transparency
while at the same time ensuring that the security principles
REQ3 demands support for heterogeneous test benches
of the original approach are not impacted.
REQ4 demands support for real and virtual equipment or
The paper is structured as follows: The next section pre-
a mix of both in a hybrid test environment
sents a requirements analysis for controlling test benches
remotely, which is based on the requirements analysis al- REQ5 demands support for cryptographically secure
ready presented for monitoring hybrid test benches in the data exchange
cloud [2]. This analysis forms the basis for the main con- REQ6 demands support for an authentication and author-
cept, which is presented subsequently, followed by a de- ization infrastructure (AAI)
scription of the prototype that was used for evaluation. Af- REQ7 demands that no LTM require a direct Internet
ter that, existing solutions and related approaches are dis- connection
cussed. The paper closes with a conclusion and outlook.
REQ8 forbids full access to the test bench
2. Requirements Analysis REQ9 demands that the mechanism not rely on VPNs or
special firewall settings (to be easily deployable in
To efficiently control a test bench, it is a prerequisite that a company environment)
its components provide information regarding their current REQ10 demands that the user interface (UI) be adapted to
state and potential errors. In the VHTNG project, these the end user’s device
components are referred to as modules. It is assumed that
The next requirements ensure that the mechanism scales
every module that is connected to the communication in-
well and is generally robust and cost-effective:
frastructure (see [3]) exposes its status monitoring interface
and its control interface. The control interface depends di- REQ11 demands that the mechanism itself not impose any
rectly on the existence of the status monitoring inter- limits on the number of test sites, test benches, and
face [4]. Consequently, remote control inherits all require- their contained entities.
ments from the remote monitoring of test benches, which REQ12 demands that the mechanism can adapt to the size
were discussed extensively by Stockmann et al. [2]. The and budget of any company
following subsection will summarize these requirements. REQ13 demands that the mechanism must support com-
putationally constrained, low-bandwidth environ-
2.1 Requirements for Remote Monitoring ments on the end user’s side
The first requirement (REQ1) specified in [2] defines the REQ14 demands that the mechanism is reliably available
minimal set of information relevant for monitoring. It con- around the clock
sists of: It must be noted that remote monitoring being the base for
1. Status information remote control is not the only reason for which all these
• Information about the type, health status, and requirements are valid. In fact, even if one were to consider
state of a module (see Figure 1) having a “blind” remote control, that is, a control without
• Information that a module is ‘alive’ monitoring, the requirements would still be relevant. For
example, control should be possible independent of the ac-
• Hierarchy of the test bench
tual location of a test site (REQ2) and it should not be re-
2. Log information quired to have the LTM exposed to the Internet to start a
Together, the status and log information represent what is simulation (REQ7). Also, a complex firewall or VPN setup
hereafter referred to as monitoring information. should not be required (REQ9) just to load a configuration.
Still, there are requirements that are control specific. These
Children are discussed in the following sub section.

* «enumeration» 2.2 Additional Requirements for Remote Control


Health
MonitoredEntity Apart from those mentioned in the previous section, there
Unspecified
are two additional requirements that are exclusive for the
+ Name: string {id} Ok
control aspect. Firstly, following VHTNG and the analysis
+ Type: TypeId Warning
+ Health: Health Error
conducted by Stockmann and Himmler [4], this paper as-
+ State: State Critical sumes that the key piece of information relevant for control
is the command. Note that in the context of a test bench,
multiple aspects can be handled using commands. This pa-
Figure 1. UML diagram of the VHTNG model con-
per, however, only considers control commands, which are
taining the essential status information (cf. [2])
used to perform certain tasks on the test equipment, such as
loading a configuration or starting a simulation.
Traditionally, commands are implemented in the form of
remote procedure calls (RPC, cf. [5]) in a request/response

ETTC 2021– European Test & Telemetry Conference


fashion. In addition, commands can have parameters to
convey extra information. Remote controlling the modules
of a test bench should “feel” just like on-site. Thus, REQ15
demands that the approach supports a command-based con-
trol paradigm.
Having a way to remotely control the modules of a test
bench always entails certain risks. No matter how well the
AAI is set up (see REQ6), accounts can become compro-
Back end
Cloud
mised or security vulnerabilities can be uncovered. As of Web server and reverse proxy
May 2021, over 1,000 entries are listed in the CVE data-
base1 in conjunction with authorization and over 1,800 en-
tries with authentication. Therefore, the mechanism must
provide a reliable means of checking and, if applicable, Get static UI Push site lo- Query
overriding/blocking the remote control (REQ16). Push cation, sta- and re-
com- Query status tus monitor- ceive
mands monitoring & ing & log com-
3. Cloud-Based Monitoring and Control Concept log updates updates mands
There are well-known traditional approaches to controlling
a test bench remotely, such as terminal-based access, for Gateway
START()
example, via secure shell (SSH) or remote desktop. They STOP()
START()
all fail to meet the requirements of status monitoring, which STOP()

has already been discussed [2]. In summary, there are three VHTNG communication infrastructure
main problems: Front end
1. They offer full access to the bench, which is not
desired as per REQ8 Test site
2. Because it is too dangerous to expose such ser-
vices directly, they must be shielded using, for ex- Figure 2: Overview of cloud-based monitoring &
ample, a virtual private network (VPN) failing to control
comply with REQ9
3. The user interface (UI) does not adapt to the end This approach also fulfills REQ16, as it can serve as the
central point to inspect incoming commands before for-
user device, failing to comply with REQ10
warding them. Also, the remote-control interface can be
It has been found that a better approach is to let the cloud fully deactivated should it be necessary by simply disabling
act as a shield between the user and the test bench (or test command processing on the gateway. Nothing needs to
site). This is also true for control. Here, it must be made change regarding the other modules.
sure that no direct connection to the bench is required To evaluate the concept, a prototype has been imple-
(cf. REQ7). mented, which is presented in the next section.
The approach presented in [2] achieves this using an Elas-
ticsearch service. It retrieves the site and status monitoring 4. Prototype and Evaluation
information as a log message from a dedicated gateway and
answers requests for status monitoring information from The prototype is based on the one presented in [2], which
end user devices. This way, the end user device never has has been extended to support the execution of commands.
a direct connection to the test bench. Likewise, the gateway It uses the same virtual machine deployed in Microsoft Az-
at the test site does not know the end user’s device. This ure2 to host Elasticsearch. Nginx3 is used to serve the static
fact can be exploited and the paradigm can be inverted to content of the frontend and to act as a reverse proxy, thus
realize the control interface. allowing both the gateway and the clients to use standard
The idea is to have the end user device and the gateway http(s) ports.
switch roles. In other words, the end user device pushes Figure 3 illustrates the current frontend, which is based on
commands (as per REQ15) in the form of a structured log a single-page application. It provides an overview of all test
message into the Elasticsearch database and the gateway sites (worldwide).
queries them. Naturally, this does not fail to meet any of
the requirements presented in section 2.1 as the mechanism
itself does not change. The only change is that the gateway
now acts as a controller (instead of just an observer) inside
the local test bench. An overview of the additions to the
original concept is depicted in Figure 2.

1 3
CVE® Program website: https://cve.mitre.org Nginx web site: https://www.nginx.com
2
Microsoft Azure Cloud website: azure.microsoft.com
ETTC 2021– European Test & Telemetry Conference
powered end user devices. The goal is to make the approach
more modular and efficient.
The gateway has been extended in two ways, as shown in
Figure 5. First, it now implements an Elasticsearch client
to query the command messages. Second, it now acts as a
VHTNG control module, so it can forward the decoded
control commands to the respective modules in the local
test bench.

Gateway Module
gRPC
Collector
and HTTP(S)
serializer
VHTNG
Adapter Elas-
START()
ticsearch STOP() HTTP(S)
client

START()
STOP()

VHTNG communication infrastructure


Figure 3: Screenshots of the remote monitoring front
end on a smartphone Figure 5: Gateway with Elasticsearch client
Here, users can hover over a test site and select a contained While evaluating the prototype, it has been found that the
test bench. This opens the test bench overlay, presenting concept works well. A major benefit is that all of the con-
the individual modules of a test bench. Users can then open trol tasks performed remotely are logged automatically.
a context menu from each of the modules, allowing them There are, however, also some minor drawbacks and chal-
to send the control commands. The frontend uses the status lenges that must be overcome in the future.
information of a module to identify which commands are
One problem is an elevated risk of a race condition due to
currently allowed according the module’s state. The other
the overall latency and loose coupling between the local
commands are disabled and consequently grayed out, as
bench, Elasticsearch, and the client device. This occurs
can be seen in Figure 4.
when a command requires a module to be in a dedicated
state. For example, consider a hypothetical STOP com-
mand that requires the targeted module to be in a
STARTED state. The frontend verifies the current state of
a module and only allows a user to send commands that are
allowed in the current state. However, it is possible that the
front end might not have received the notification that a
module state has changed (e.g., in this case, the module
may have changed to a STOPPED state due to an error).
The front end consequently allows the user to send com-
mands that are in fact prohibited for the actual (local) state.
This results in the command seemingly having no effect,
apart from an error, which can lead to confusion. However,
it must be noted that, in principle, this issue can also appear
locally, although it is more unlikely due to the smaller la-
tencies.
Figure 4: Screenshots of the context menu containing
currently available control commands All in all, the prototype shows that the approach is viable,
that is, the monitoring approach originally presented in [2]
The control command itself is encoded into a structured log can be extended to support the remote control of test
message. It includes the site name, the command name, the benches without violating any of the security-related re-
target, and a timestamp when the command was sent. The quirements. The next section discusses some related ap-
log message is pushed into the Elasticsearch database in the proaches and solutions.
cloud. In the current version of the prototype, the client de-
vice generates the required Elasticsearch queries directly, 5. Related Work
effectively bypassing the back end depicted at the top left
of Figure 2. In a future version, the back end will provide a The concept of monitoring and controlling test benches re-
simple, more convenient API that is better suited for low- motely is still a rather new topic in the aerospace industry

ETTC 2021– European Test & Telemetry Conference


and is slowly being addressed by test equipment manufac- transmit the data to a server in the cloud, the data must first
turers. be translated to MQTT or another IoT middleware. Any
There are numerous solutions on the market that make it control commands sent to a machine via the cloud must
possible to monitor and evaluate test data remotely. But first be translated from an IoT middleware command to a
currently, there are no solutions available that fulfill all the command that can be understood by the OPC UA client.
requirements presented in section 2. Using an IoT middleware usually requires all clients to be
SystemLink4 is a solution from National Instruments for directly connected to the Internet, which violates REQ7.
connecting, deploying, and managing test benches online. There are ways to mitigate that, such as using a gateway as
For the purposes of SystemLink, managing test systems a proxy, but this would still result in a more complex solu-
means installing software packages and rebooting devices. tion, compared to our log message-based approach.
This is possible via a Salt5 interface. The Salt interface can
be used to manage the IT infrastructure, even from third- 6. Conclusion and Outlook
party vendors. To manage the systems via Salt, additional
This paper has presented a solution that allows users to
software must be installed on each device. However, it is
monitor and control test sites anywhere on the globe. It uses
not possible to configure or reconfigure a test bench with
the same architecture and technology stack as a cloud-
different test configurations and to control the states of the
based monitoring approach that was presented recently.
LTM or the test via control commands, like on-site. This
fails to comply with REQ15. Thereby, it also fulfils all the strict requirements regarding
security and cost-effectiveness. In contrast to existing ap-
For remote monitoring, SystemLink uses the IoT middle- proaches, the test equipment does not require a direct con-
ware AMQP. The monitored data is transmitted directly to nection to the Internet and there is no way an adversary
a server from each LTM. A server provides the transmitted could gain full access to the test bench. The control aspect
data to users via a web application. The server can be set has been designed to fit perfectly in the log-based nature of
up on-site or in the cloud. In contrast to the presented ap- the monitoring approach by encoding the control com-
proach, it is necessary to install additional software on each mands into log messages, thereby persisting them in the log
LTM, which must then also be connected directly to the database. The on-site gateway is the single consumer of
Internet. This fails to comply with REQ7 and requires the these control commands and could be easily deactivated in
firewall setting to be adjusted, thus failing to comply with case of a security leak, so local testing can continue.
REQ9. Furthermore, in order to use SystemLink in a vir-
tual and hybrid test bench (see REQ4) consisting of sys- The presented prototype proves that the approach can be
tems from different suppliers (see REQ3), those systems implemented in the cloud, although this is not strictly nec-
essary, and it can be deployed in a self-hosted environment
must implement the non-standardized SystemLink inter-
just as well, thus providing maximum flexibility.
face, resulting in a vendor lock-in.
There are still some features missing though. For example,
Other solutions include PAK cloud6 from Müller-BBM,
it is not currently possible to easily load a test bench con-
Diagnostics 4.07 from Softing, and Lab Discover from
Keysight Technologies8. All of these systems claim to al- figuration. This is because the frontend has no means of
low for remote monitoring of test benches. With some so- querying the available configurations. One future task
therefore is to find a suitable protocol to convey this infor-
lutions, it is possible to define notifications. However, it
mation.
does not seem to be possible to control the systems and in-
tervene in the tests remotely (failing to comply with Another open issue is the back-end API to decouple the
REQ15). Furthermore, it remains unclear whether they can front end from the Elasticsearch database and allow for
be used with devices from different manufacturers. This more bandwidth-efficient communication.
fails to comply with REQ3. Finally, with regard to Elasticsearch, it is necessary to eval-
In other industries, there are cloud-based solutions for con- uate whether the recent changes in licensing10 impact the
trolling industrial machines and sensors. applicability of the approach and whether a migration to a
fork is required.
One approach described by H. Raju et al [6] is based on
OPC UA9 [7]. An OPC UA server collects the data from
the industrial devices and sensors (OPC UA client). To

4
National Instruments Corp., "SystemLink 18.2 Manual": files/pdf/ae/poster/WEB_180831_1_softing_Poster_Diagn
https://www.ni.com/documentation/en/system- ostics.pdf (Accessed May 2021)
8
link/18.2/manual/manual-overview/ (Accessed May 2021) Keysight Technologies: "Scienlab Lab Discover,"
5
Salt: "SALTSTACK Salt Documentation": https://www.keysight.com/de/de/assets/7018-06287/bro-
https://docs.saltproject.io/en/pdf/Salt-3003.pdf (Accessed chures/5992-3238.pdf (Accessed May 2021)
9
May 2021) OPC UA: https://opcfoundation.org/
6 10
PAK cloud website: https://www.mbbm- FAQ on 2021 License Change for Elasticsearch:
vas.com/en/products/data-management/pak-cloud (Ac- https://www.elastic.co/pricing/faq/licensing (Accessed
cessed May 2021) May 2021)
7
Diagnostics 4.0 Poster: https://automotive.soft-
ing.com/fileadmin/sof-
ETTC 2021– European Test & Telemetry Conference
7. Acknowledgement 9. Glossary

Part of this work was funded by the German Federal Min- AAI: Authentication and authorization infrastructure
istry of Economic Affairs and Energy. – provides support for authentication and au-
thorization
AMQP: Advanced Message Queuing Protocol – an open
standard for message transmission based on
middleware
API: Application Programming Interface
HTTP: Hypertext Transfer Protocol – a common appli-
cation layer protocol
HTTPS: A secure variant of HTTP using Secure Sockets
Layer (SSL) encryption
IoT: Internet of Things
LTM: Laboratory test means - a single piece of test
equipment that is part of the test environment in
a test bench
8. References MQTT: Message Queue Telemetry Transport - a proto-
col standard for publish/subscribe-based mid-
dleware
[1] D. H. Martinen, M. Lagalaye, J. Pfefferkorn and J. OPC UA: Open Platform Communications Unified Archi-
Casteres, "Modular and Open Test Bench Architecture tecture – an open standard, developed by the
for Distributed Testing," SAE International, Hartford, OPC Foundation which defines a protocol for
Connecticut, 2017. machine-to-machine communication
[2] L. Stockmann, F. Nolte and S. Reglitz, "Monitoring of UI: User interface
Virtual and Hybrid Test Benches," in SAE VHTNG: Virtual and Hybrid Testing Next Generation – a
International, Orlando, Florida, 2021. research project that aims to create a generic and
[3] A. Himmler, L. Stockmann and D. Holler, modular architecture for an open test environ-
"Communication Infrastructure for Hybrid Test ment
Systems - Demands, Options, and Current VPN: Virtual Private Network – a network, which can-
Discussions," SAE Int. J. Aerosp., vol. 9, no. 1, pp. not access from third users
134-139, 2016.
[4] L. Stockmann and A. Himmler, "Monitoring and
Control of Hybrid Test Systems," SAE International,
Hartford, Connecticut, 2017.
[5] J. E. White, "RFC 707: A High-Level Framework for
Network-Based Resource Sharing," in Proceedings of
the 1976 National Computer Conference, New York,
1976.
[6] H. S. R. a. S. Shenoy, "Real-time remote monitoring
and operation of industrial devices using IoT and
cloud," in 2016 2nd International Conference on
Contemporary Computing and Informatics (IC3I),
Amity University, Noida, India, Dec. 2016.
[7] Fraunhofer IOSB-INA, "Industrie 4.0
Kommunikation mit OPC UA Leitfaden zur
Einführung in den Mittelstand," Forum Industrie 4.0,
Fraunhofer-Anwendungszentrum Industrial
Automation (IOSB-INA), Frankfurt am Main, Lemgo,
2017.

ETTC 2021– European Test & Telemetry Conference

You might also like