Quality of Service (QoS) for Edge
We are at an interesting juncture for Edge services getting platforms to become more intelligent! One recent example of this is Google’s TPU (https://cloud.google.com/edge-tpu/) on edge giving us capability to do AI! Other great examples are – AWS Greengrass (https://aws.amazon.com/greengrass/) enabling us to run compute, caching, ML capabilities for edge side computing, as well as Azure IoT Edge (https://azure.microsoft.com/en-in/services/iot-edge/) providing comprehensive Intelligent Edge! We have been witnessing trends, advancements & implementations in IoT happening with family of Smart * (Things) like Smart Home, Building, City, Healthcare, Infrastructure – for Personal & Commercial segments as well as Smart Manufacturing with Industrial IoT (IIoT) segment. This is growing bigger in size, faster than anticipated thereby leading to data proliferation as well as narrowing the gap between physical & digital world. In this regard, there is certainly a need to ensure measuring & retaining quality for such Edge services which can be defined through apt SLAs.
Such SLAs would be specific to industry & respective problems being solved. Here, depending on industry, below notions can be existing, applicable to Edge Services:
1) Soft deadline: Deadline which if missed won’t cause any fatal/critical issues
2) Hard deadline: Deadline that cannot be missed which can cause fatal/critical issues
3) Overrun allowed: Time taken to respond by the Edge service after deadline which is allowed
4) Slack time: Time between deadline & remaining time of execution for respective Edge service
Also similarly, outcome from such Edge Services would vary based on industry & respective problems being solved.
So for example for Healthcare, in case of patient monitoring involving human life, it can have stringent SLA to meet (High) i.e. absolutely, deadline in terms of response time of Edge service cannot be breached – so candidate for Hard deadline. For Smart Building involving optimized usage/reduced cost use-case can be moderate SLA to meet (Medium) i.e. it might be OK to miss the deadline if it’s still within the budget – so candidate for Semi-Hard or Soft depending on how much Slack existed (becomes Semi-Hard case i.e. overshooting budget)/overrun allowed (becomes Soft case i.e. still within budget). For Home entertainment systems use-case, it can have comparatively lower SLA to meet (Low) i.e. perfectly OK to miss the deadline for genuine reasons like misunderstanding of voice recognition leading to unable to play a song – so probable candidate for Soft deadline.
Now the question is – Can we quantify this Quality for Edge Services which will enable us to define right SLAs & perform better SLA management? Answer is – Yes!
This quantification of Quality for Edge can come from QoS-E (Quality of Service for Edge) which can be defined as a combination of some function of Quality of Experience for Edge (QoE-E) & Quality of Perception for Edge (QoP-E).
QoS-E = f (QoE-E) + f (QoP-E)
Many interesting questions arise here beginning with –
1) What is QoE-E & QoP-E?
2) How to compute QoE-E & QoP-E?
3) Can we automate QoS-E by automating QoE-E & QoP-E?
4) What is the manual seed needed for QoE-E or QoP-E?
5) What it takes to automate QoE-E & QoP-E?
6) How to define the functional relationship with QoE-E & QoP-E?
7) How can we derive SLAs from this?
8) Can we apply this uniformly to variety of Edge services within specific industry?
9) How can we adaptively apply it to various Edge services across industries?
10) Can we make SLA management autonomous based on QoE-E?
QoS can be defined & achieved to position one’s Edge service(s). This is important enabling standardization for customers.