Service Metrics vs. Service Measures in Hybrid Cloud
By Clinton France | Feb 07, 2019
Our performance engineers at Krystallize frequently field questions about the difference between CloudQoS®-enabled Service Measures and traditional operational Service Metrics furnished by Cloud or infrastructure monitoring systems. The answer has to do with perspective.
Service Metrics track utilization, IO, and other system activities within a discrete operating environment. Service Measures however test and quantify actual end to end throughput, productivity or performance of the operating environment ability to deliver on a application demand (workload profile). A simple way of thinking about metrics vs. measure is looking at the dashboard in the car. Tachometer (RPMs), Water Temperature, Oil Pressure, tire preassure, Gas Gauge tell you how different components in the car are running. However, the speedometer tells you how productive the car is at delivering you from point a to point b. (miles per hour is an outcome mesure based on taking a workload from place at to be with a given driving pattern).
When applying this concept to Cloud services, you might think of traditional Service Metrics as a set of competent metrics (similar to the Engine RPMs, Transmission gear, etc) on the overall operation of systems. Where Service Measures go one step further measuring the total performance potential of the Cloud service when driving a given ‘workload profile’ (application demand).
What yesterday’s traditional operational metrics and monitoring systems do very well is collect, correlate and display component data. What is does not do is test and assess the total performance potential of that operating environment to deliver on your workload. In other words, you can have two identical systems with what appears to be identical Service Metrics AND they can have DRASTICALLY different Service Measures.
In the real world example above, you can see that even in Private Cloud this is an issue: The CloudQoS® Service Measures on two Private Cloud instances using the Same Private Cloud Architecture, Same hardware, Same Software and Same Service Operator – showed over 70% different Service Measures
In follow-on blogs, I will share more leanings from the above and also outline why this so important within today’s hybird Cloud operations. However, I will leave you with the following link for a paper published by AWS VP’s Adrian Cockcroft (Cloud Architect and visionary) during his tenure at Netflix Cloud architect: “Utilization is virtually useless as a metric!” http://www.hpts.ws/papers/2007/Cockcroft_CMG06-utilization.pdf
Simply put, while operational metrics are helpful, they are essentially meaningless without a true understanding of your Cloud’s ability to deliver (Capability to meet your workload demand). CloudQoS® sets the service bar by which operational Cloud metrics make sense.