SLALOM: a Language for SLA specification and monitoring

IT services provisioning is usually underpinned by service level agreements (SLAs), aimed at guaranteeing services quality. However, there is a gap between the customer perspective (business oriented) and that of the service provider (implementation oriented) that becomes more evident while defining and monitoring SLAs. This paper proposes a domain specific language (SLA Language for specificatiOn and Monitoring - SLALOM) to bridge the previous gap. The first step in SLALOM creation was factoring out common concepts, by composing the BPMN metamodel with that of the SLA life cycle, as described in ITIL. The derived metamodel expresses the SLALOM abstract syntax model. The second step was to write concrete syntaxes targeting different aims, such as SLA representation in process models. An example of SLALOM’s concrete syntax model instantiation for an IT service sup-ported by self-service financial terminals is presented.

💡 Research Summary

The paper addresses a long‑standing problem in IT service management: the disconnect between the business‑oriented view of customers and the implementation‑oriented view of service providers when defining and monitoring Service Level Agreements (SLAs). To bridge this gap, the authors propose a domain‑specific language called SLALOM (SLA Language for Specification and Monitoring). The core idea is to model SLAs and the underlying business processes within a single, formally defined meta‑model, thereby enabling consistent specification, validation, and monitoring across stakeholder groups.

The development of SLALOM proceeds in two major steps. First, the authors identify the essential concepts required to describe an SLA life‑cycle as defined by ITIL‑v3 (contract creation, negotiation, execution, review, termination) and the elements of a BPMN (Business Process Model and Notation) process model (activities, events, gateways, pools/lanes, data objects). Both sets of concepts are expressed as UML class diagrams enriched with OCL (Object Constraint Language) constraints that capture static semantics—for example, a metric assigned to a service must match the metric defined for the service’s goal. By applying a meta‑model composition technique (equivalence operator), the two meta‑models are merged into a unified abstract syntax model. This unified model resides at the M2 layer of the Model‑Driven Architecture (MDA) stack and serves as the formal grammar of the SLALOM DSL.

The second step defines concrete syntaxes that map the abstract model to usable representations. Two concrete syntaxes are provided: a graphical syntax that extends BPMN diagrams with SLA annotations (e.g., attaching “availability ≥ 99.99 %” or “response‑time ≤ 2 s” to tasks, and linking penalties to violation paths) and a textual DSL syntax that uses declarative statements together with OCL constraints to describe contracts, goals, metrics, data sources, and measurement methods. Both syntaxes are automatically checked against the abstract syntax model, guaranteeing well‑formedness and semantic consistency before deployment.

To illustrate the approach, the authors present a case study involving self‑service financial terminals such as ATMs and POS devices. The service is modeled as a BPMN process, and SLA clauses are embedded directly into the process model using SLALOM’s graphical extensions. Metrics like transaction success rate, response time, and error rates are linked to concrete data sources (log files, network monitors, application traces). The meta‑model captures the relationships among services, goals, metrics, measurement activities, and data sources, enabling the automatic generation of monitoring pipelines that can evaluate SLA compliance in near real‑time.

Key contributions of the paper include: (1) a unified meta‑model that integrates SLA concepts with BPMN process concepts, eliminating the semantic gap between business and technical perspectives; (2) the use of OCL‑based static validation to detect inconsistencies at the model level; (3) dual concrete syntaxes that cater to both business analysts (graphical) and system engineers (textual); and (4) a demonstrable application to a realistic financial‑service scenario.

The authors acknowledge limitations such as the increased complexity introduced by meta‑model composition and the lack of an automated runtime bridge that continuously synchronizes operational monitoring data with the SLALOM models. Future work is outlined to address these issues through model‑to‑model and model‑to‑text transformations, integration with cloud‑based monitoring platforms, and the automation of the full SLA life‑cycle—including automatic generation of corrective workflows upon SLA violations.

In summary, SLALOM offers a formal, model‑driven solution for specifying and monitoring SLAs within IT service environments, promising improved alignment between customers and providers, early detection of contract breaches, and a foundation for automated, end‑to‑end service level management.