Towards a Service Management Information Base Martin Sailer Munich Network Management Team University of Munich (LMU) Oettingenstr. 67, D-80538 Munich, Germany martin.sailer@mnm-team.org Abstract. The trend towards service-oriented computing has implications on the area of IT management. While traditionally management of networks and sys- tems has dominated research as well as industrial applications, the focus is now turning to service management. One of the key concepts of integrated manage- ment is the management information base (MIB), the composite of management information. While a number of standardized MIBs have been devised in the area of network and systems management, a comprehensive MIB for service man- agement is still missing. The PhD thesis presented in this paper aims at filling this gap by specifying semantic-rich and business-oriented service management information. 1 Introduction With service-oriented computing becoming widely accepted by the industry, the number of commercial deployments is steadily increasing. The success of these deployments largely depends on the providers’ ability to efficiently manage their IT infrastructure. To meet this requirement service providers deploy a wide array of hardware and software to deliver and manage services to their customers. Each management application deals with a certain facet of service management and therefore uses valuable data as key to its operation. The resulting situation is that information about services needs to be widely and electronically shared across departments throughout the provider’s enterprise - and with partners and suppliers. Such information is part of a Management Information Base (MIB), one of the key concepts of integrated management. The MIB contains the definition of the entities that are managed (managed objects), including their properties, their permissible behavior, and the operations that may be performed upon them [1]. Properties of managed objects are explicitly described by attributes, that usually consist of types and values. A MIB enables integration of different management applications in that it allows for exchange of semantically bound management information. For this reason, a large number of standardized managed object (MO) definitions have been devised in the area of network and systems management. However, a comprehensive MIB for service management is still missing. Hence, the realization of interoperability is impeded by the lack of a common information base for service management. While a number of modeling frameworks for service management information have been devised, they currently show deficiencies regarding description of common service properties. The PhD thesis presented in this paper aims at addressing this gap with 26 Martin Sailer the concept of a Service Management Information Base (SMIB). The SMIB approach aims at specifying semantic-rich, business-oriented service management information. This includes common service management attributes as well as association information between services and resources. In Section 2 a typical service provisioning scenario is presented in order to derive requirements for a Service MIB. Related work in this area is discussed in Section 3 and juxtaposed to the identified requirements. Section 4 introduces the SMIB approach and presents the further steps to be taken in the PhD thesis. 2 Scenario The Web Hosting Service offered by the Leibniz Supercomputing Center (LRZ) serves as an example IT service (see Figure 1). The LRZ operates the sci- entific network in Munich and hosts web sites for other research institu- tions. Figure 1 shows a simplified view of the service incorporating three Loadb main components: an array of web routers servers, a global filesystem (AFS), and a database. In addition, network com- ponents need to be considered for man- agement (Loadb routers, Nx-switches). Nx-switches The management systems respectively MIBs employed by the LRZ provide means to manage networks and appli- cations, but do not take into account services or dependencies between ser- vices and resources.Yet, that knowl- edge has to be deduced by systems administrators and maintained sepa- rately. It therefore often lacks formal- ization and is usually not reusable. Webservers Thus, a comprehensive service man- agement repository is required to en- able consistent management. The chal- lenges stemming from that requirement AFS-Filesystem Database can be further categorized into specifi- Fig. 1. Web Hosting Service at the LRZ cation of service attributes and formal- ization of service resource dependencies. 2.1 Specification of meaningful service attributes Traditionally, management information has been defined from a vendors’ perspective and was therefore often restricted to information that could be easily obtained by instru- menting components [2]. With the shift to service management, that perspective has to Towards a Service Management Information Base 27 be adapted in a way that it reflects the providers’ information needs. Furthermore, ser- vices are managed along their lifecycle, in different functional management areas, thus motivating the inclusion of these aspects. Additionally, service provisioning – and con- sequently its management – may cross administrative boundaries. Hence, specification of service management information needs to be tailored towards the following generic requirements: Alignment to providers’ informa- tion needs The MIBs currently em- Ma pi nf o B us ine ployed contain information pertain- rm Mg s s at i mt. SLA ing to applications and network el- on Se ements. Examples for such data are Ma r vi ce Service nag MIB the number of dropped IP pack- Ma em ent pi ets, the CPU load of a webserver, nf o Ap rm pli c atio or http-requests per second. While ati nM on ana gem MIB this may suffice from a network ent tion Ne lica t wo rk a App IB and systems management point of Ma n rk M nag d Sys wo view, service management requires em t em Net ent s abstracted and enriched data that reflects the providers’ information needs. That is, information needed Fig. 2. Management Pyramid to deliver the service according to the objectives set up in service level agreements (SLAs).This includes in- formation relating to the quality level of the service, to the customers having subscribed to the service, or the impact of service failures on the business. Administrative domains Application scenarios like dynamic service provisioning or Grid computing require services to be composed in a highly dynamic and flexible man- ner, thereby crossing organizational boundaries. Accordingly, the ServiceMIB must provide capabilites to encompass different management domains and therefore allow to present different views on service management information according to the needs of the involved organizations. The functional areas of management, usually classified by the OSI FCAPS, dissect the whole complex of management. Yet, management information is shared between different functional areas, e.g., service usage information is a prerequisite for account- ing as well as fault management. Consequently, all of these areas should be included in a ServiceMIB that purports to deal with service management. The service lifecycle usually consists of a planning, negotiation, provisioning, opera- tion/change, and withdrawal phase. Since management information is propagated and enriched among these phases, the ServiceMIB must accommodate all phases. 2.2 Formalization of dependencies between services and resources Service management relies on knowledge regarding dependencies between services and resources. In particular, dependencies are the key to service problem resolution in that 28 Martin Sailer they provide a foundation for determining causality [3]. Formalization of dependen- cies has to accommodate different levels (e.g., functional, organizational) as well as provide means to express complex dependencies (e.g., dependencies causing certain service quality degradations). Therefore, information contained in system, network and application MIBs needs to be combined in such a way that it reflects the service. Conse- quently, a formalization of dependencies must allow for translating/mapping informa- tion from other MIBs to leverage an integrated management approach (see Fig. 2). 3 Related Work In this section prominent representative work in the area of management information modeling is presented and its significance in the focus of this paper is discussed. Common Information Model (CIM) The Common Information Model [4] introduced by the Distributed Management Task Force is an object-oriented information model that aims at providing a common way to represent information about networks and systems as well as services. The CIM Core Model contains the specification of the class service, which acts as a superclass from which specific service classes (e.g. print service) can be derived. Additionally, dependencies between services and other CIM elements (e.g., ServiceAffectsElement) are expressed by association classes. However, only few service attributes have been defined so far for the basic service class as well as associations per- taining to that class. CIM constitutes an comprehensive modeling framework that could be extended to meet the requirements delineated in Section 2. Since it accomodates network, application, and service management information, it offers a foundation for modeling dependencies between those areas. Web Services Distributed Management Two complementary specifications have been presented by the Oasis Web Services Distributed Management technical committee: Management Using Web Services (MUWS) deals with providing a common manage- ment interface for distributed IT resources, whereas Management of Web Services (MOWS) introduces the application of the former onto Web Services [5,6]. Service Management information is exposed through manageability capabilities and catego- rized into common and resource-specific capabilities. At present, only basic capabilites like the CurrentOperationalState, the LastResponseTime or NumberOfRequests have been defined. Moreover, the WSDM approach does not explicitly address dependencies between services and resources. NGOSS SID As part of the New Generation Operations Systems and Software (NGOSS) program the TeleManagement Forum released a Shared Information/Data (SID) model [7]. It employs an object-oriented modeling approach and defines a frame- work for defining service management information from a business-oriented point of view. SID offers a promising modeling framework for service management information that builds upon sound design principles. However, it also shows deficits regarding the description of common service attributes. Internet Information Model The IETF model revolves traditionally around the Simple Network Management Protocol (SNMP) [8]. This is a read/write, state based protocol for altering variables in a management information base (MIB) and is used by the vast Towards a Service Management Information Base 29 majority of commercial products. Despite of newly introduced RFCs that deal with the specification of service management information for WWW services[9], the Internet Information model does not offer object-oriented modeling features, and is therefore not suitable for modeling of service management information. 4 Approach The Service MIB approach aims at specifying semantic-rich, business-oriented service attributes as well as association information between service and resources. It therefore consists of two main steps, that are aligned with the challenges presented in Section 2: Specification of business-oriented service attributes In recent time, service manage- ment has been influenced by the strong trend towards process-orientation. Process- oriented frameworks contribute to that trend in that they provide guidance for enter- prises striving to deploy IT service management processes. A prominent example for such a framework is the IT Infrastructure Library (ITIL). ITIL has been introduced by the British Office of Government Commerce (OGC) and provides a collection of best practices for IT processes in the area of IT service management [10,11]. Service management is described by 11 modules which are grouped into Service Support Set and Service Delivery Set. These modules describe processes along with its goals and activities as well as necessary databases and interfaces. To facilitate the definition of service attributes we leverage a use-case driven ap- proach. The main idea is to analyze ITIL service management processes in order to derive common service attributes. Since such processes essentially represent a specifi- cation for service management tasks, they provide a comprehensive collection of use cases that can be further utilized to derive service management information needs. This approach offers considerable benefits with respect to the requirements denoted in Sec- tion 1. Since ITIL constitutes a business-driven approach to IT service management, it already takes lifecyle and interorganizational aspects into account, while at the same time offering a sufficient level of genericity. Example: The Incident Management Process deals with restoring normal service operation. One of the core activities within this process is the classification of incidents, consisting of e.g., priorization of incidents. To support this activity, service management information relating to the business impact of the incident is required. That is, the SLA associated with the service, the impact on related services, or affected users. Formalization of service-resource dependencies Formalization of association infor- mation regarding services and resources requires high expressive power that cannot be expressed by single attributes. Instead, a more complex formalism needs to be employed that allows to specify associations on a per-attribute basis in a functional manner. This includes means to map information defined for network and systems to ServiceMIB attributes. Therefore, this issue has been approached with a declarative language to specify service-resource dependencies, called Service Information Specification Lan- guage (SISL). This concept has been presented in [12], which introduced a monitoring architecture that facilitates the creation of service management information from raw monitoring data according to a SISL specification. 30 Martin Sailer Further Work Previous work in the context of the incident and problem management processes has analyzed use-cases in the area of service-oriented event correlation and impact analysis [3]. This work has shown applicability of the approach and will be extended towards other ITIL processes. While up to now, basic examples of SISL for expressing the op- erational state of services have been drafted [12], further work on that topic will target modeling of more complex scenarios. Acknowledgment The author wishes to thank the members of the Munich Network Management (MNM) Team for helpful discussions and valuable comments on previous versions of this paper. The MNM Team directed by Prof. Dr. Heinz-Gerd Hegering is a group of researchers of the University of Munich, the Munich University of Technology, and the Leibniz Supercomputing Center of the Bavarian Academy of Sciences. Its web–server is located at http://www.mnm-team.org. References 1. Black, U.: Network Management Standards - The OSI, SNMP and CMOL Protocols. McGraw-Hill (1992) 2. Hegering, H.G., Abeck, S., Neumair, B.: Integrated Management of Networked Systems – Concepts, Architectures and their Operational Application. Morgan Kaufmann Publishers, ISBN 1-55860-571-1 (1999) 651 p. 3. Hanemann, A., Sailer, M., Schmitz, D.: Assured Service Quality by Improved Fault Man- agement — Service–Oriented Event Correlation. In: Proceedings of the 2nd International Conference on Service–Oriented Computing (ICSOC04), New York City, NY, USA, ACM SIGSOFT and SIGWEB, ACM Press (2004) 183–192 4. Distributed Management Task Force (DMTF): Common Information Model (CIM) Version 2.9. Specification (2005) 5. Organization for the Advancement of Structured Information Standards (OASIS): Web Ser- vices Distributed Management: Management Using Web Services (MUWS 1.0) Part 1 (2004) 6. Organization for the Advancement of Structured Information Standards (OASIS): Web Ser- vices Distributed Management: Management of Web Services (WSDM-MOWS 1.0) (2004) 7. TeleManagementForum: Shared Information/Data (SID) Model Concepts, Principles, and Domains. Technical Report GB 922 Member Evaluation Version 3.1 (2003) 8. Case, J., McCloghrie, K., Rose, M., Waldbusser, S.: RFC 1902: Structure of management information for version 2 of the simple network management protocol (snmpv2). RFC, IETF (1996) 9. Hazewinkel, H., Kalbfleisch, C., Schoenwaelder, J.: RFC 2594: Definitions of managed objects for www services. RFC, IETF (1999) 10. Office of Government Commerce (OGC), ed.: Service Support. IT Infrastructure Library (ITIL). The Stationary Office, Norwich, UK (2000) 11. Office of Government Commerce (OGC), ed.: Service Delivery. IT Infrastructure Library (ITIL). The Stationary Office, Norwich, UK (2001) 12. Danciu, V., Sailer, M.: A monitoring architecture supporting service management data com- position. In: Proceedings of the 12th Annual Workshop of HP OpenView University Asso- ciation. Number 972–9171–48–3, Porto, Portugal, HP (2005) 393–396