Service Procurement Scenario - NEXOF-RA - BT

From FutureInternetWiki

Contents 1 Service Procurement 2 What If Questions 3 Research Challenges 4 Opportunities for FIRE Facilities Use

[edit] Service Procurement

In a future service ecosystem there will exist providers and consumers for services at all level, ranging from complete business processes at the top through to simple hardware resources at the bottom. A consumer will identify services within the structure of their business processes, that after careful consideration of the transaction costs and other factors such a security, reliability etc, they decide to purchase from an external supplier. The economist Ronald Coase initially proposed the concept of transaction costs in 1937 in his seminal paper “The nature of the Firm”, they are widely used to assess when an external service is more cost effective for a business than developing the same functionality internally.

Having decided on the need to purchase a service externally, the potential consumer has to formulate a description of the service they require. This description needs to include both functional and non-functional requirements that the customer has identified for the service and be compatible with a search technology.

The requirements are input into a search engine that may include some semantic intelligence to identify equivalent services, a list of suitable suppliers and then their service offerings are returned to the consumer.

The consumer is then able to choose a suitable service and supplier based on both the offered functional properties of the service and the context for the service as described by the non-functional properties that would include details such as price, confidence in supplier etc.

Having chosen a supplier, the consumer then requests details of the available quality levels that the service provider guarantees for the service and will complete some form of SLA template with the quality levels they think most closely match their requirements.

The completed SLA template is passed to the supplier to obtain agreement for the use of the service and after confirming the start of the agreement the service provider will send the consumer a service end point reference so that the consumer can begin to use the service.

During consumption of the service both service consumer and service provider will collect and analyse performance information. This is to ensure that the service both meets the customers’ requirements, and that the quality guarantees embodied in the SLA are being maintained.

Either during use of the service, or at the end of the period of use, the consumer may have to make payment for their use of the service. The size of the payment and payment mechanism will have been defined in the SLA agreed between consumer and provider before use of the service began.

The rationale for this scenario is the following: in a modern, flexible, service oriented world companies are free to focus on their core business and increasingly are using specialist providers to provide key parts of their business processes. The advantages of such an approach includes flexibility, reduced costs (particularly in pay per use environments) and the ability to switch between suppliers as the need arises, to benefit from either better pricing or better service.

To support a market of this type there is a requirement for some consistency in service description to support discovery, comparison and potential switching between suppliers. This scenario applies to:

• Any consumer of any service
• Any provider of services: the service provider needs to advertise the service they are offering, define the service levels available and provide SLA templates to potential purchasers of the service.
• Any service integrator/developer: the service and any metrics used to assess its performance need to be described in a consistent and meaningful fashion.

There are a number of challenges that need to be addressed in order to fully realize the scenario described:

• The scenario relies upon common understandings between the service consumer and service provider of details of the services functional and non-functional characteristics. This may require the development of standardized vocabularies for such descriptions.
• The process of describing and then searching for a service may require the assistance of semantically enriched search and discovery tools so that all appropriate candidate services are correctly identified.
• When the service consumer requests SLA templates they must have an understanding of the terms embodied in the SLA, and relationships between them, leading to a requirement for standardized SLA representations.
• Should the SLA be ambiguous or not exactly what the service consumer requires there may be a requirement for a negotiation protocol for the agreement of SLAs. This is particularly complex because the service provider must have an adequate management structure in place to accommodate any changes in the base SLA.

Both parties in the agreement should be able to access, if required, meaningful and consistent performance data for the service, to assist in management of the SLA. There are a number of architectural constraints highlighted by the described scenario:

• A uniform representation is required for services of different types. These services will be offered depending on customer demand and not on any arbitrary definition of functional layers in the architecture. Many services will span a number of traditional technology layers.
• There is a requirement for a decentralized architecture, but that will support federation, for directories, SLA repositories etc. each of which may have different policies governing access and use, to support public and private service offerings.
• Service discovery mechanisms need to address both functional and non-functional characteristics of offered services. They also need to accommodate selective control of access and visibility of services.

There needs to be standard representations of SLAs, and standard protocols for their negotiation and monitoring.

[edit] What If Questions

1. What if standard representations of SLAs, and standard protocols for their negotiation and monitoring or uniform descriptions for services of different types are not taken up by a large community?
2. What if many devices/channels do exist, on which the services will be created or consumed? (mobile devices, television, telephone, washing machine, refrigerator, printer, sensors) which are sometimes switched off?
3. What if consumers are not able to generate their requirements?
4. What if consumers are not able to select services?
5. What if providers cannot fill in the SLA templates of their services?
6. What if it is not possible to find SLA descriptions of complex composed services across providers? Especially if compositions are generated on-the-fly?
7. What if several industries provide services in incompatible ways (like in the socio-economics scenario nr 3. What if there are many internets? (Telecom providers that provide services only accessible through their own network, media providers doing the same on their broadband network, IT services on the internet).

[edit] Research Challenges

1. What if standard representations of SLAs, and standard protocols for their negotiation and monitoring or uniform descriptions for services of different types are not taken up by a large community?

• a. This is a big risk. Developing standards is one way to approach the problem. But at the same time, the take up of the standard should be achieved. Since projects are working now on developing the necessary input for standards bodies, short term research in 2015 could aim at applying the standards at large scale and see if the above scenario is realistic in a convergent world.

2. What if many devices/channels do exist, on which the services will be created or consumed? (mobile devices, television, telephone, washing machine, refrigerator, printer, sensors) which are sometimes switched off?

• a. There will exist many devices/channels on which services will be created or consumed. Projects now are working on the development of tools which facilitate the creation of services on mobile phones, or easy mashup of services on the desktop of computers by non-IT experts. Research for 2015 could concentrate on mechanisms to make service creation easier and more intuitive, for non-professionals as well as for professionals. Also they should try to reconcile the more formal service world (between businesses and with strong SLA's) with the informal service world, so that both worlds can benefit from each other.

3. What if consumers are not able to generate their requirements?

• a. This is a genuine risk. It is well-known in software engineering that users are not good in stating requirements which have the level of granularity necessary for software development. Therefore research could focus on enabling consumers to develop good requirements (here I think of the "early service discovery" work in SeCSE and the software requirements research), or semi-automatic requirement generation, which requires (long term) research on linking service consumer profiles and contextual information coming from IoT and devices to requirements.

4. What if consumers are not able to select services?

• a. Too much choice is also not desirable. Based on user preferences a preselection can be made automatically and the highly ranked services can be presented in order of relevance to the consumers.

5. What if providers cannot fill in the SLA templates of their services?

• a. The project SLA@SOI is addressing this at the moment. They will develop mechanisms to derive SLA's both analytically (by analysing code and resource consumption) and empirically by monitoring and observing to fill SLA templates. In 2015 these mechanisms can be tried out on a large scale and be integrated with a global service delivery platform.

6. What if it is not possible to find SLA descriptions of complex composed services across providers? Especially if compositions are generated on-the-fly?

• a. This is a longer term research issue. SLA@SOI will start to work on this, but to have proper SLA descriptions of on-the-fly composed services will be very difficult. Especially to predict performance or to reconcile different payment models (for instance if one service has a 1 year subscription based payment and the composed service will be pay per use, how much of the payment needs to go to the subscription service?). What if the composed service is to use part of a movie in a different situation? Is it necessary to pay for the whole movie, or just a part of it?

7. What if several industries provide services in incompatible ways (like in the socio-economics scenario nr 3. What if there are many internets? (Telecom providers that provide services only accessible through their own network, media providers doing the same on their broadband network, IT services on the internet).

• a. To have many incompatible internets is an undesirable situation. However, the possibility exists. Therefore, the different industries that are converging (media, telecom, IT) should work together and understand the opportunities they can have together. This can be achieved by testing convergent services on the FIRE facilities.

[edit] Opportunities for FIRE Facilities Use

Through FIRE it could be checked whether services provided by telecom operators, media providers and IT providers can be made fully interoperable, so that it will be possible to compose these types of services to new ones, tailored to the needs of users. Important to check whether all players will receive the payments they are entitled to and whether individuals, SME's or other organisations are able to build new convergent services that are taken up by end users. Can we find the right type of SLA description and service description and the right balance between openness and control that is necessary for stakeholders to become part of such a service procurement market, including individuals that use their mobile devices or other devices to adapt services to their needs?