Tel: +49 711 459 22895, Fax: +49 711 459 22961, Email: {karaenke | kirn}@uni-hohenheim.de. Abstract: The .... For further applicability of automatic processing ...
Service Level Agreements: An Evaluation from a Business Application Perspective Paul KARÄNKE, Stefan KIRN University of Hohenheim, Information Systems II, 70599 Stuttgart, Germany Tel: +49 711 459 22895, Fax: +49 711 459 22961, Email: {karaenke | kirn}@uni-hohenheim.de Abstract: The vision of a worldwide computing grid evokes the scenario of electronic markets for grid services. For such a market, explicit formal statements of both obligations and guarantees regarding grid services in a business relationship are needed; these are referred to as service level agreements (SLAs). The objects of our investigation are current specifications and standards for SLAs. We address the problem that these approaches do not sufficiently fulfil requirements of business transactions, because they have been designed for other, technical objectives. We evaluate the most relevant SLA specifications and standards from an e-business transaction perspective, clarifying how they currently cover business transactions on e-markets, in particular which activities within market transactions they support. We assess the current state of SLA approaches and reveal room for possible enhancements.
1.
Introduction
Grid technology is an emerging field and potential use cases are broadly based [14]. We anticipate that developments towards electronic markets for grid services will achieve reality in the foreseeable future. These markets will be different from existing grid application approaches. Instead of mid- or long-term relationships formed manually, they will provide means for automatically matching service providers and requesters. Explicit formal statements of obligations and guarantees regarding grid services in a business relationship are often referred to as service level agreements (SLAs) [15, pp. 1-5]. A SLA provides the operational definition of a service as part of a contract between a provider and a consumer. Grid-related SLA approaches aim to provide a service abstraction while facilitating measurement and monitoring of agreed service properties [3, pp. 6-7]. This paper deals with existing specifications and standards for SLAs such as WSLA [6] and WS-Agreement [1]. In markets with prevailing short-term business relationships, increased standardisation and automatisation offer potential for lowering transaction costs. Therefore, we imagine advanced SLA approaches as an integral part of a comprehensive ebusiness-oriented framework supporting grid markets. The problem, however, is that current SLA approaches are not sufficient to support such electronic markets. These specifications have been designed for the purpose of describing technical features of services in a machine-readable and unambiguous way. However, current SLA approaches do hardly cater for business-oriented aspects of a contract such as pricing and negotiation. In this research, we address this problem from an e-business transaction perspective, thus we analyze whether and how transactions on electronic markets for grid services can be supported by existing SLA concepts and respective specifications. In particular, we work towards a clarification of the question to which degree relevant activities in each phase of a market transaction can be covered by current SLA approaches.
Management science and information systems provide a comprehensive repertory of models and methods on e-business transactions (e.g., [13][4]). We adopt a mature ebusiness transaction model which was also employed by other researchers (e.g., [5]). In section 2, we describe the e-business transaction model that will be further used in our analysis and evaluation. In section 3, we select the most relevant SLA specifications. Then, we apply our set of criteria to these approaches in section 4. Finally, section 5 discusses our findings and points out to future avenues of research.
2.
E-Business Transactions
Transactions on traditional and electronic markets correlate with a finite number of interactions between market participants. Classes of grouped interactions form phases of a transaction [13, p. 1]. The sub-processes within such a phase will hereafter be referred to as activities. As a starting point, we focus on a dyadic interaction between a customer and a supplier as suggested in [4, p. 2]. We extract the activities of encapsulated interactions for both customer and supplier as described in the model by Schmid and Lindemann [13]. The original model contains two more activities which precede the browse activity, namely gathering information concerning the general business environment and technology. In the following, we do not consider these activities as they are not directly related to a particular service and therewith SLA. Table 1 shows the results from the elucidated considerations. Table 1: Phases of a Market Transaction (following [13])
Name Information phase
Definition Gathering information about (i) market participants and (ii) goods and services.
Activities B – browse: search for market partners O – offer: make an offer
Result List of suppliers/ customers
Agreement phase
Negotiations between customer and supplier concerning a matching offer revealed in the information phase. The parties agree on mutual commitments.
N – negotiation: determine service properties and price
Contract
Service delivery and payment; all necessary operations agreed upon are performed.
D – delivery: delivery of goods and/or services P – payment: payment of goods and/or services
Settlement phase
3.
A – agreement: conclude a contract Fulfilment of mutual commitments
SLA Specifications
In this section, we select the most relevant SLA specifications for the grid area. We select WSLA and WS-Agreement, since these approaches are the most widely accepted ones. Additionally, we select a semantic extension of WS-Agreement, as it is a very promising approach, aiming towards mitigating points of criticism [12]. Further, it exemplifies the assets and drawbacks of WS-Agreement’s flexibility in comparison. Finally, it preserves compliance to the original specification. All three approaches have a XML-based syntax. 3.1
Web Service Level Agreements (WSLA)
Web Service Level Agreements in terms of IBM’s WSLA specification [6] contain the parties involved, the service definitions as well as guarantees with references to the latter. The main focus lies on performance constraints [2, p. 2]. The descriptions of the parties involved consist of exactly one service provider and exactly one service consumer
(signatory parties). In addition, an arbitrary number of supporting parties which provide services to the signatory parties but cannot be made accountable with reference to the SLA, can be described (e.g., measurement service providers). Service definitions contain descriptions of the provided functional operations, observable properties relevant for the SLA, metrics (i.e., functions) for these properties, and measurements for the metrics. Further, WSLA allows defining aggregation functions, which operate on low-level parameters. Obligations, also referred to as guarantees, comprise of particular explicit duties which have to be completed by a certain party with regards to the parameters defined in the service definitions. Additionally, execution of actions under well-defined circumstances can be made obligatory. Although the WSLA specification primarily focuses on Web Service (WS) descriptions and a special extension for WSDL references exists, it is conceptually not limited to WS descriptions and can be used for describing virtually any kind of service [6, p. 22]. 3.2
Web Services Agreement (WS-Agreement)
WS-Agreement is expected to become an Open Grid Forum (OGF) recommendation. This paper refers to the latest specification draft published by the OGF for public comments [1]. The WS-Agreement specification defines a format for agreement descriptions, a simple protocol for establishing agreements, and interfaces to monitor agreement parameters on run time. WS-Agreement includes information about the parties and terms in a “context” associated with the agreement. A service term can be either a non-standardised description, a reference to a service, or a description of observable properties. The latter are used to define service level objectives (SLOs) in guarantee terms, similar to WSLA. The strength of WS-Agreement lies in its extensibility: it assumes that the metric language is domain-dependent. Therefore, the language is not specified, which may be regarded as lack of structure [9, p. 2], but offers increased flexibility. WS-Agreement also allows assigning a set of domain-dependent “business values” to each guarantee in terms of service provision (SLOs) such as importance, cost, or penalties. The WS-Agreement specification offers a basic mechanism for negotiation, though it is not significantly versatile [12, p. 2-3]; in particular, an offer can only be accepted or rejected [1, p. 37]. 3.3
WS-Agreement with SWAPS Extension
Details on the proposal for a semantic extension of WS-Agreement can be found in [9]. As this approach comprises an extension of WS-Agreement, we only review the enhancements. The approach aims at facilitating partner selection (i.e., matching of capabilities and requirements) by semantic annotations of the WS-Agreement definitions. It comprises annotations with tags which link expression parameters to ontological concepts to mitigate ambiguities and provide semantic concept mapping. For further applicability of automatic processing and reasoning, the potential lack of structure of WS-Agreement is avoided by introducing several tags of WSLA (e.g., parameter) as well as additional custom tags (e.g., unit), while preserving much of WS-Agreement’s flexibility.
4.
Coverage of E-Business Transactions
In this section, we evaluate the selected SLA specifications with regard to the e-business transaction model, and assess the degree of coverage and flexibility regarding each activity; e.g., we gather information each specification includes about negotiations and compare it to the set of negotiation types such as auctions. In case of a distinct activity and specification, we evaluate extensibility of the approach regarding a certain activity; e.g., if and how it is possible to negotiate SLAs beyond the scope of a specification. We review corresponding literature and give references where appropriate, substantiating our assessments.
4.1
Information Phase
The information phase concerns gathering information about market participants and services. Within the browse activity, a participant searches for market partners meeting predefined functional & non-functional service properties requirements. This can comprise any guaranteed technical property (e.g., availability) or business aspect (e.g., price). WSLA supports browsing for a service to a limited degree, as its concepts comprise only a small subset of all functional and technical non-functional properties. Economic attributes such as cost model can not be represented. The assignment of “business values” to SLOs in WS-Agreement and its flexible structure facilitate individual extensions and therewith improve potential applicability, making non-standard additions a necessity. Current approaches for WSLA matching are purely syntactic [9, p. 9], because there are no means for explicit semantics in WSLA. A matching algorithm that compares the single nodes of a WSLA tree is presented in [16]. The WS-Agreement specification does not support explicit semantics either. As syntactic matching is a challenge with regard to even marginal differences, WSLA and WS-Agreement are considered to offer only elementary support for the browse activity [9, pp. 1-2]. The main objective of the semantic SWAPS extension to WS-Agreement is advanced support for matching of capabilities and requirements. Consequently, support for browsing activities is facilitated (cf. section 3.3). However, no approach enables coverage of all functional and non-functional attributes; e.g., nothing can be stated about service security. WSLA does not offer an explicit mechanism for making an offer. Though, published resources’ SLOs may be regarded as providers’ offers to potential customers [16, p. 3]. WS-Agreement and therewith the SWAPS extension explicitly consider an offer as a basis for negotiation [1, pp. 8, 37 et seq.]. 4.2
Agreement Phase
After receiving an offer, supplier and customer may negotiate service properties and the pricing scheme. The parties agree on mutual commitments by the result of a (i) fixed price order, (ii) auction, (iii) tender, (vi) exchange, or (v) bilateral negotiation. Any pricing scheme is regarded as a type of price (e.g., metered-usage pricing). Although explicitly stated, WS-Agreement does not offer significant support for negotiations beyond an offer, followed by acceptance or rejection [1, p. 37]. Most negotiation protocols (e.g., auctions) are incongruous with this committed step [12, p. 2][10, p. 1]. The SWAPS extension only comprises contents of agreements and therefore negotiations; hence the protocol compatibility is not altered. The WSLA specification does not include information about negotiation, though it is possible to negotiate the content of a WSLA beyond the scope of the specification. Regarding the agreement activity, WSLA and WS-Agreement target different objectives. WSLA has a more fixed structure, preferring standardisation over flexibility. As WS-Agreement leaves large parts undefined towards broader applicability, this may be regarded as lack of structure [9, p. 2]. The SWAPS extension of WS-Agreement is a hybrid of the former two approaches: towards a more structured approach, it includes structure characteristics of WSLA besides custom tags. Additionally, the semantic assignment mitigates ambiguity of agreement expressions. Agreements on performance constraints are clearly the strength of the approaches, since this has been one of the main design objectives. 4.3
Settlement Phase
The settlement phase comprises service delivery and payment, i.e., performing all necessary operations agreed upon. This phase includes provisioning of access to a service interface as well as ensuring compliance to any guarantee. The payment activity may include
determination of the final price on basis of a cost model agreed upon, e.g., consumptionbased pricing on a fixed cost model basis. All approaches provide means for monitoring delivery in the scope of the agreement at least for performance constraints. However, even if the customer trusts the monitoring provider, the WSLA model also implies a trust relationship between the monitoring service and the provider, regarding the delivered low-level values, and thus does not enforce monitoring [7, p. 2]. WS-Agreement and the SWAPS extension delegate the trust to the language used to describe the guarantees [7, p. 2][8, p. 2]. Although the approaches include extensive support for monitoring deliveries, security considerations (e.g., authentication and authorisation) are not being addressed [1, p. 49]. In WSLA, business values including payment information cannot be linked to raw and aggregated values [8, p. 2]. WS-Agreement and its extension include the “business value” field, though they do not include explicit accounting and billing mechanisms. 4.4 Summary Towards the objective of performance constraints, WSLA offers sufficient means for defining and monitoring performance properties. Other issues are virtually not addressed. WS-Agreement enhances the set of representable contents, though it comes with the cost of leaving large parts undefined. The SWAPS extension mitigates some of the points of criticism; though lack of support is especially noticeable in economic and negotiation regards. Table 2 summarises the preceding assessments. (+ : covered
Activity Browse
Offer Negotiation
Agreement
Delivery Payment
Table 2: Evaluation Matrix ○ : partly covered - : not covered)
Sub criteria
WSLA
WS-Agreement
WS-Agreement/ SWAPS
functional properties performance properties other technical nonfunctional properties economic properties supplier’s offer customer’s offer
+ ○
○ + ○
○ + ○
-
○
○
-
+ +
+ +
fixed price order auction tender
-
+ -
+ -
○
○
exchange bilateral negotiation
-
-
-
-
○
○
functional properties performance properties other technical nonfunctional properties economic properties
-
○
○
+
+
+
○
○
○
+ -
○
○
+
+
○
○
-
5.
Discussion and Outlook
The main contribution of this research is a clarification of the question to which degree activities of market transactions are being covered by current SLA approaches relevant for the grid area. The evaluation has shown that current SLA approaches offer differentiated support for transactions on e-markets. As mentioned, we did not evaluate the approaches with regard to their original objectives, thus any lack of support for e-business transactions must not be interpreted as general criticism. The most obvious lack of support concerns economic attributes as well as negotiation protocols. Additionally, other non-functional properties (e.g., security) may have to be improved by additional concepts to make the approaches more suitable for business transactions. Room for possible enhancements does also exist in the agreement activity. With regard to this issue, Rule-Based Service Level Agreements (RBSLA) [11], which aims at unconstrained expressiveness and flexibility for SLAs, may come to attention. Due to integrating rules from Horn Logic, Event-Condition-Action (ECA), Event Calculus, Deontic Logic, Courteous Logic, and Description Logic, RBSLA offers extremely rich expressiveness for virtually arbitrary SLA rules, based on RuleML. Further, it allows procedural attachments (i.e., written in programming languages) for delegating computingintense tasks as well as integrating external data sources via nesting of SQL and XML/RDF queries. The foreseen unconstrained expressiveness and flexibility and the applied method make this approach very promising, though the complexity may mitigate applicability; in addition, matching of policies is out of the approach’s scope [9, p. 9]. The complexity may also be a reason for the absence of projects utilising RBSLA. Future research may include further studies of the identified lack of coverage. Additionally, an advanced transaction model may be required to further represent special requirements on grid markets. Contrarily, existing approaches already provide support for activities which are not covered by the selected model; e.g., dispute activities for conflict resolution can be facilitated utilising the concept of third, external monitoring parties. Addressing the problem of insufficient support of electronic markets by current SLA approaches, the SWAPS extension of WS-Agreement offers possibilities beyond its original goals. References to ontological concepts may also facilitate interpretation in negotiations. Improvements could be achieved by further developing this approach with regard to business application requirements, requiring modifications and extensions of the simple protocol for establishing agreements and extensions for economic aspect representation. Other relevant research topics include but are not limited to: general terms and conditions, IPR and licensing issues, scheduling, and accounting and billing.
Acknowledgements This work has been supported by the BREIN project (http://www.gridsforbusiness.eu) and has been partly funded by the European Commission’s IST activity of the 6th Framework Programme under contract number 034556. This paper expresses the opinions of the authors and not necessarily those of the European Commission. The European Commission is not liable for any use that may be made of the information contained in this paper.
References [1] [2]
Alain Andrieux, Karl Czajkowski, Asit Dan, Kate Keahey, Heiko Ludwig, Toshiyuki Nakata, Jim Pruyne, John Rofrano, Steve Tuecke, and Ming Xu. Web services agreement specification (WSAgreement), proposed OGF recommendation. Open Grid Forum, Sep 2006. Gustav Boström, Pablo Giambiagi, and Tomas Olsson. Quality of service evaluation in virtual organizations using SLAs. In Proceedings of 1st International Workshop on Interoperability Solutions
[3] [4] [5]
[6] [7]
[8]
[9] [10] [11]
[12] [13] [14] [15] [16]
to Trust, Security, Policies and QoS for Enhanced Enterprise Systems (IS-TSPQ 2006), page 12, Bordeaux, France, 2006. Karl Czajkowski, Ian Foster, and Carl Kesselman. Grids in context. In Ian Foster and Carl Kesselman, editors, The Grid 2: Blueprint for a New Computing Infrastructure, chapter 1, pages 259–283. Morgan Kaufmann Publishers Inc., 2004. G. Goldkuhl and M. Lind. Developing e-interactions—a framework for business capabilities and exchanges. In Proceedings of the 12th European Conference on Information Systems (ECIS2004), Turku, Finland, 2004. Steffen Lamparter, Stefan Luckner, and Sibylle Mutschler. Formal specification of web service contracts for automated contracting and monitoring. In Proceedings of the 40th Annual Hawaii International Conference on System Sciences (HICCS'07), Waikoloa, Big Island, Hawaii, USA. IEEE Computer Society, 2007. Heiko Ludwig, Alexander Keller, Asit Dan, Richard P. King, and Richard Franck. Web service level agreement (WSLA) language specification. IBM Corp., 2003. Olivier Nano and Mark Gilbert. SLA monitoring: Shifting the trust. In Proceedings of the eChallenges e-2005 Conference, Ljubljana, Slovenia - Innovation and the Knowledge Economy: Issues, Applications, Case Studies, Paul Cunningham and Miriam Cunningham (Eds), 2005 IOS Press Amsterdam, ISBN: 158603-563-0. Olivier Nano and Marcel Tilly. Filling the gap between SLA and monitoring. In Proceedings of the eChallenges e-2006 Conference, Barcelona, Spain - Exploiting the Knowledge Economy: Issues, Applications, Case Studies, Paul Cunningham and Miriam Cunningham (Eds), IOS Press, 2006 Amsterdam, ISBN: 1-58603-682-3. Nicole Oldham, Kunal Verma, Amit Sheth, and Farshad Hakimpour. Semantic WS-Agreement partner selection. In Proceedings of the 15th international conference on World Wide Web (WWW'06), Edinburgh, Scotland, pages 697–706. ACM Press, 2006. Michael Parkin, Dean Kuo, and John Brooke. Framework & negotiation protocol for service contracts. In Proceedings of the IEEE International Conference on Services Computing (SCC'06), Chicago, Illinois, USA. IEEE Computer Society, 2006. Adrian Paschke. RBSLA a declarative rule-based service level agreement language based on RuleML. In Proceedings of the International Conference on Computational Intelligence for Modelling, Control and Automation and International Conference on Intelligent Agents, Web Technologies and Internet Commerce Vol-2 (CIMCA-IAWTIC'06), Sydney, Australia, pages 308–314. IEEE Computer Society, 2005. Shamimabi Paurobally and Michael Wooldridge. A critical analysis of the WS-Agreement specification. Technical report, Department of Computer Science, University of Liverpool, 2005. Beat F. Schmid and Markus A. Lindemann. Elements of a reference model for electronic markets. In Proceedings of the 31st Annual Hawaii International Conference on System Sciences (HICSS'98), Waikoloa, Big Island, Hawaii, USA, volume 4, page 193. IEEE Computer Society, 1998. Larry Smarr. Resource and service management. In Ian Foster and Carl Kesselman, editors, The Grid 2: Blueprint for a New Computing Infrastructure, chapter 1, pages 3–12. Morgan Kaufmann Publishers Inc., 2004. Dinesh Verma. Supporting Service Level Agreements on IP Networks. Macmillan Technical Publishing, 1999. Weilai Yang, Heiko Ludwig, and Asit Dan. Compatibility analysis of WSLA service level objectives in self-managing systems. In DSMS - IEEE First Workshop on the Design of Self-Managing Systems. IEEE, 2003.
