NMRG/EMANICS Workshop Position Statement Stefan Wallin Data Ductus 1 Service Quality 1.1 Who we are We work as network management integrators for Telco Operators world-wide. The HP OpenView products are the main solution components. We have been working with classical alarm management and lately service management and SLA management. We do part-time research work at CDT Luleå University. 1.2 Introduction We will look at the concept of service quality within the telecom service provider context. The subject of service quality is very complex and broad in its nature. The two terms service and quality are inherently lacking solid definitions. Adding the two into a new concept does not make the picture more clear. We will use the two following definitions: * Service: Something that the service consumer pays a service provider for. * Quality: A measure correlated to customer satisfaction with the service. An underlying complexity is the different contexts of a service; one as seen by the service provider and on the other hand as seen by the service user: * Delivered Service Quality: often an objective measurement of the service quality using technical performance indicators. The time window is often short if not even momentarily * Perceived Service Quality: a subjective judgement of the service quality. Covers a brode aspect of service quality indicators over a long period of time; Installation of the service, contacts with helpdesk, etc. 1.3 Different ways of measuring service quality There are several different approaches to measure service quality, each of them with focus on one, or both, of the above contexts. * Traffic shaping: this is a technology which tries to manage the network service in order to actually deliver the defined QoS. This is mainly the efforts of the IETF work on QoS. Different techniques are applied to manage throughput, packet-loss, latency and jitter. Well-known IETF work is Diff-Serv and MPLS. * Probes: simulates end-users behaviour using the network as a black-box in order to measure the actual delivered network service quality. Cisco Service Assurance Agent, SAA, falls into this category. Mobile networks have probes for probing telecom network services like Voice, MMS, etc.. * Modelling, calculation: a system where a formal model ties different measurements into a service model. The service quality is actually calculated by a calculation engine. These are fairly complex tools where a formal language is used to define the service model. The model captures the overall structure of the services, and mapping in several layers down to individual network resources. It calculates so called Key Performance/Quality Indicators, KPIs/PQIs, using expressions based on input events and polls. Service state is propagated in the service model which is typically tree-structured. Since there is a model the system and user can perform 'reasoning', a certain service state can be reduced to the original events/polls. What-if scenarios can also be applied, showing the affected services as effects of network events and changes. * User feed-back systems: informal methods like web forms, telephone interviews etc in order to catch users perception of the service Apart from covering different contexts they also have different capabilities: * Monitor: only reporting on real-time and historical QoS * Control: can actually control network resources in order to deliver a defined QoS * Root-cause: can drill down from a service failure to the root-cause of the failure 1.4 Organisational aspects It is impossible to talk about service quality without looking into the involved parties. * Service user: context of perceived service quality. * Network provider: network management center, network operations centre, etc: context of delivered service quality. The NMC needs to have an understanding from network events/state into affected services and customers in order to prioritize work. * Customer care: needs to map the two contexts to each other: o Customers are complaining, why? o Which services and customers will be affected by a network failure? 1.5 Pros and cons of different techniques Traffic shaping + Actually controls the network to deliver service quality + Embedded in some protocols - Only available for some network services Probes + Easy to deploy + Measures the service usage - No understanding of why?, root-cause etc. Modelling, calculation + Reasoning: service-impact, root-cause, what-if + Creates a model of the network - Complex to deploy - QoS defined by calculations, they may be wrong User feed-back + Reports on true user perception + Easy to capture - No mapping to why? - Subjective 1.6 What is service quality, really? Although tools and standards for QoS have been available for a while. We are not sure that users of services perceives a big improvement. The number of available services have increased but the service quality have rather decreased. What is the reason? We think that one of the problems is that the QoS concepts have been focusing on the provider side and the technical/characteristics side of quality. We need to add the perceived quality as defined by subjective measurements. Furthermore, delivered and perceived service quality needs to be integrated. User | User Terminal | Service Access Point | Media | Service Delivery Point | Provider infrastructure | Service producer The above ASCII art illustrates the chain involved in the total service usage chain. Our proposed overall architecture is following: * Service model which models the total chain above o BUT, the service model needs to be dynamic and easy to change o The model needs to support pivoting in order to present different views * Collection of quality parameters from all components in the chain, both technical provider and user subjective indicators.