A formal approach to Information Lifecycle Management.
Lars Arne Turczyk, Oliver Heckmann, Rainer Berbner, Ralf Steinmetz · TUbilio (Technical University of Darmstadt) · 2006
In this paper we present a framework for simulating Information Lifecycle Management (ILM) scenarios. The framework is derived from a formal approach to ILM which offers concrete mathematical terminology. INTRODUCTION Information Lifecycle Management (ILM) is one of the great trends in the context of information storage. It has its roots in hierarchical storage management (HSM), which was popularized with mainframe storage management strategies in the early 1980s. ILM is a process for managing information through its lifecycle, from conception until disposal, in a manner that optimizes storage and access at the lowest cost. ILM is based on the idea that in an enterprise there are different information with different values. The different information will be stored on different storage devices. The same principle is employed with ILM for storage systems. ILM manages information according to its value. Valuable information is stored on systems with high Quality of Service (QoS) [1]. The value changes over time and therefore migration of information to cheaper storage systems with lower QoS is required. Automated migration makes ILM dynamic. In this paper we build a framework for testing the quality of migration rules by simulation. SNIA’S DEFINITION OF ILM ILM as a concept is not easy to handle. Therefore in 2004 the Storage Networking Industry Association (SNIA) gave a new generally accepted definition [2]: Definition 1 (ILM SNIA) Information Lifecycle Management is compromised of the policies, processes, practices, and tools used to align the business value of information with the most appropriate and cost effective IT infrastructure from the time information is conceived through its final disposition. Information is aligned with business processes through management policies and service levels associated with applications, metadata, information and data. This definition forms the basis for an accurate occupation with ILM. Nevertheless it is general and has limitations when applied to specific cases. Therefore a formal approach will lead to specific results for the employment of ILM as we will show. The following definition is not meant as a contradiction to the SNIA definition. It aims to help creating a best practices framework for ILM and ILM evaluations. FORMALIZED DEFINITION OF ILM The formalized definition of ILM is derived canonically. To get a common understanding we refer to SNIA’s definition of information [2]: Definition 2 (Information) Information is data that is exchanged, expressed or represented within a context such as an application or a process. This means the application offers the context for data. Information in ILM has a granularity and the number of information in an enterprise is finite. An access is based on operational reasons. This shows that the information has a certain importance for business. This importance is defined as value of the information. Definition 3 (Value of Information) The value of information V(I) describes the importance of the information I for the business. The value of information can be expressed in money. V(I) is initiated with the creation of the information I. The time of creation is defined as t 0 ≥ 0. The value of an information changes over time. It is a function of t: V(I(t)). Therefore it is necessary to migrate the information during its lifecycle to adequate storage systems. How is the value determined? This is a complex question in ILM. The easy answer is: It depends on the the business processes. To be more specific it comes, for example, from the administrator, the end user or the CIO or from the application. External regulations and laws can determine the value, too, e.g. Sarbanes Oxley Act (SOA). Furthermore the value can be derived from the usage of information. Files with many accesses are more valuable than files with few accesses, which are more valuable than files with no accesses. Therefore observing access patterns is one way to determine the value. There is a long tradition in looking for access patterns [3][4]. In section 4 we refer to our own study, which was conducted in 2005 [5]. When the value is determined, the information are grouped into information classes according to their value. Values change over time, so the constellation within an information class varies. It is dynamic. One strength of ILM is to take this dynamic into account. Definition 4 (Information Class) An information class C is a set of all information I 1 ,..., I m , whose values V(I i (t)) lie at the time t in a predefined (value-)interval. b a b t I V a t I C C j i j i j i , ; )) ( ( ) ( : : , An information class is a set of information which have similar values. Different information classes have different (disjunct) intervals. The intention is to store the content of an information class on the same type of storage devices. IDEA GROUP PUBLISHING This paper appears in the book, Emerging Trends and Challenges in Information Technology Management, Volume 1 and Volume 2 edited by Mehdi Khosrow-Pour © 2006, Idea Group Inc. 701 E. Chocolate Avenue, Suite 200, Hershey PA 17033-1240, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com ITB12752 532 2006 IRMA International Conference Copyright © 2006, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. Definition 5 (Storage Class) A storage class S is a set of storage devices with similar properties, i.e. Quality of Service (QoS) and cost. QoS summarizes especially security, backup frequency, access speed [6]. The storage classes represent the hierarchies in the ILM solution. The content of an information class is not static. The value of each information might change beyond the value interval [a,b). Then an information becomes an element of a different information class which is mapped to a different storage class. The information will be migrated. That is the migrating process of ILM. The changes in value of an information are dynamic and define a “lifecycle”. The following definition puts the lifecycle in a formula. Definition 6 (Lifecycle) Let 0 ≤ t 1 ≤ t 2 . The lifecycle L of information I is the mapping of the value of I between time t 1 and t 2 . 2 1 )) ( ( : ) ( t t t t I V I L The change in the value of information during a period of time represents a lifecycle. The lifecycle is dynamic for each information. The question of administrators is “What happens to the system when all the dynamic lifecycles are reflected to the storage environment?”. To avoid a bad experience on the real storage environment the dynamic behaviour of ILM has to be simulated. Definition 7 (ILM) Information Lifecycle Management (ILM) is the mapping of the information I 1 ,..., I n on Classes C 1 ,..., C m , according to their values V(I 1 ),..., V(I n ) in time interval [t 1 , t 2 ]. APPLICATION OF THE FORMAL APPROACH ILM is a dynamic process affecting the whole IT. The effect of an ILM solution depends on the quality of the migration rules. To improve the quality of the rules simulations are helpful. The formal definition offers the framework for simulations. Step 1: V(I), the value of information I is determined. Step 2: I becomes element of an information class according to its value. Step 3: The determined information class has a fixed related storage class. Step 4: I is stored on a storage device related to the determined storage class. What is a migration rule exactly? A migration rule takes the value of an information and determines whether a migration is to be executed or not. In the framework the migration rule is a part of the information class. By defining the value interval [a,b) of an information class the migration rule is defined, too. Therefore the migration rule is part of determinating the value, too. The more complex the rule, the more complex the metric of the value and vice versa. Again, valuation is not easy. The framework can handle both simple and complex valuations. Case Study In a case study on a database we provided following results [6]: There were more than 150,000 files on the system and 89 percent of them were not accessed 90 days after creation. The intention is to create an ILM concept with three different storage classes. The files shall be migrated between the three hierarchies automatically. The task is to find an adequate and efficient set of migration rules. To give an impression how specific rules can be evaluated we show the process for testing the rule “Move information when it is not accessed for a period of time (90 and 200 days)”. Application of the formal approach to a case study The framework with its steps 1-4 is amended by step 0, the assessment, and by step 5 the simulation. Specific Case: Step 0: There are 150,000 files from different office applications on a database.