Quality of Service must be Measurable
Quality of service (QoS) is a success measure. QoS is a measure of performance but like most measures, your metrics must be relevant.
QoS has been a measure of services such as network performance for some time. Traditionally QoS is a binary pass or fail per unit measure. Data points might include errors, collisions, data rate. QoS is an approach for dealing with a single problem or problem domain.
People have different expectations related to QoS based on their experience and expectations. People also consider multiple factors when determining overall performance of a device or solution. For example, people consider a vehicle as good or bad based on many criteria in different ways, many non tangible.
A security by design approach, as used in Beacon’s “No Trust” solution, views QoS as a multi faceted measurement. For example, IoT implementation includes end devices, network interfaces and software functions.
Telemetry data sources provide data used for measurement. Measurement effectiveness is based on information quality, quantity, diversity or source and relationship.
By comparing data sources, one is able to paint a more accurate QoS picture based on multiple measures with variable outcomes. Variable measures provide degrees of effectiveness. Security and QoS is enhanced when using multiple data sources and relationship hierarchy.
QoS is built upon: QoS to Requirements, QoS to Perfection and System integrity System integrity includes operational components, telemetry validation, support services and System Security Security is part of a systems integrity therefore to perform effective security one needs to have measurable data related to multiple components in the relationship matrix.
QoS to Perfection = What is the best possible QoS in the system and how close are we to that measurement? (See Image1)
QoS must be part of system design to include data quality, quantity and reliability for accurate evaluation and prediction. Edge devices deliver real time telemetry, senor information, network connection information and functional information is important for effective issue identification related to multiple areas of performance evaluation. Edge periodic or sampling telemetry does not provide a reliable picture of operations. Contiguous timeline of telemetry is required to establish operational baselines. Periodic or sampling may be necessary for validation of data integrity.
Analysing information with the appropriate AI solutions automates alerts and notifications. Determining a business’s ideal outcome is the basis for determining a measure for QoS.
Telemetry from one type of sensor or source type yields operational information relevant to a narrow evaluation. It is necessary to combine data sets from multiple related sources for improved evaluation and conclusion.
Consider a simple set of temperature readings. A device with one temperature sensor provides less reliable and potentially accurate information than multiple embedded sensors. Multiple sensors allow you to compare readings to determine accuracy and potentially identify a specific issue.
Additional external environmental temperature sensors provide additional information to gain perspective related to what environmental conditions are affecting a device.
Other telemetry sources deliver telemetry which paints a larger picture related to conditions and activities that further impact a device and related operations. For example, a fan in a ventilation system has a number of temperature sensors related to motor temperature, fan speed and electrical load. Plenum temp and air flow sensors provide additional information related to external telemetry readings. By analysing the readings of these related telemetry outputs, you can determine if there is a dirty air filter and perhaps location. If the air filter is replaced and there is no change in the readings, additional actions may be required related to sensors and devices.
QoS solutions require quality telemetry. The higher the quantity of telemetry the more accurate is the analysis of telemetry. Real time and continuous data is more accurate and makes analysis more reliable.
QoS metrics, in an IoT environment, includes an array of data sets to be analysed for accuracy and relevance to produce a usable operational metric. This operational metric is compared with an ideal outcome. The result is an accurate measure for QoS.
QoS is not a simple task!
The goal is to improve monitoring, management, operations and support of IoT devices. The challenge is how to measure QoS. Different people have different expectations related to the definition of QoS for different reasons such as precision measurement. Determining QoS is not a simple task when dealing with different organizations, information types, different needs and complex data interactions.
For example, two cameras are mounted in a facility: one at the front door of a business and another around the corner monitoring the parking garage. Both cameras have a similar processing lag or delay. This delay becomes more important related to the front door camera because it is used for realtime ID validation so the delay becomes a problem. Both cameras have similar performance but the one is less acceptable than the other. QoS is based on need, not simple device operation.
Telemetry includes physical telemetry as in temperature, speed and pressure. Functional performance such as was a command successful may be equally important Did a command get sent to start a motor? Did the motor receive the command? Did the motor turn on/off?
QoS by design means data collection, analysis and response. The question is, “How do you implement an integrated solution for manage telemetry, operational and historic information to ensure QoS?”