Introduction

The measurement of properties of cloud services, especially for the purpose of cloud service level agreements (SLAs), presents many challenges, which inhibit the uptake of cloud services and inhibit the overall effectiveness of the cloud services marketplace. Metrics in practice are usually described using natural languages, typically in ‘plain English’, which is often difficult to understand, compare, and implement. Such definitions of metrics lead to many problems. Typical concerns include:

— Clarity: The metric definition may be incomplete, ambiguous, illogical, self-contradictory, or not defined at all. For example, cases exist where ‘availability’ is defined in ways which have little to do with generally accepted definitions of ‘availability’; where the definition is such that the service can be unavailable for the majority of the time yet the metric will show 100 % availability; where the metric requires continuous monitoring, which is actually not possible; or where the provider is able to determine at its sole discretion what the result is.
— Comparability: It may be impractical or effectively impossible to compare different services in terms of their promised service levels because of the significant inconsistency in how their respective metrics and SLOs/SQOs are defined.
— Implementation: It may be impractical or even impossible to measure the metric in practice, and to determine whether promised service levels have been met or not.

This document has been developed to help address these and similar concerns. It includes technical content, but the high-level concepts are expected to be understandable by non-technical individuals who understand the business context for metrics. It provides a metric model that defines the conditions and rules for performing a measurement and understanding the result.

A metric complying with the model defined by this document addresses the concerns above:

— Clarity: A definition of a metric eliminates the ambiguities which currently exist in natural language descriptions.
— Comparability: The structured nature of the metric facilitates the comparison of different metrics and SLOs/SQOs based on a metric.
— Implementation: The structured representation of the information needed to measure a characteristic facilitates the process of developing measurement tools. Likewise, if the metric is found not to be implementable, then the metric will need to be revised so that it can be implemented, and the structure of the technical specification is expected to facilitate this revision process.

The focus of this document is on metrics for cloud SLAs, but it is also usable for cloud service metrics (CSMs) that are not included in cloud SLAs [such as ones used by cloud service providers (CSPs) for their internal performance monitoring], and may also be usable for non-CSMs.

BS ISO/IEC 19086-2:2018+A1:2023

Cloud computing. Service level agreement (SLA) framework - Metric model

Current

Published:

28 Feb 2023