A distributed object database stores objects persistently at servers. Applications run on client machines, fetching objects into a client-side cache of objects. If fetching and cache management are done in terms of objects, rather than fixed-size units such as pages, three problems must be solved: 1. which objects to prefetch, 2. how to translate, or swizzle, inter-object references when they are fetched from server to client, and 3. which objects to displace from the cache. This thesis reports the results of experiments to test various solutions to these problems. The experiments use the runtime system of the Thor distributed object database and benchmarks adapted from the Wisconsin OO7 benchmark suite. The thesis establishes the following points: 1. For plausible workloads involving some amount of object fetching, the prefetching policy is likely to have more impact on performance than swizzling policy or cache management policy. 2. A simple breadth-first prefetcher can have performa...

We describe an efficient server-based algorithm for garbage collecting object-oriented databases in a workstation/server environment. The algorithm is incremental and runs concurrently with client transactions, however, it does not hold any locks on data and does not require callbacks to clients. It is fault tolerant, but performs very little logging. The algorithm has been designed to be integrated into existing OODB systems, and therefore it works with standard implementation techniques such as two-phase locking and write-ahead-logging. In addition, it supports client-server performance optimizations such as client caching and flexible management of client buffers. We describe an implementation of the algorithm in the EXODUS storage manager and present results from an initial performance study of the implementation. These results demonstrate that the introduction of the garbage collector adds minimal overhead to client operations. 1 Introduction A primary strength of Object Oriented...

In the following document we describe the current implementation of ADAMS, a parallel object oriented database system developed at the University of Virginia. The parallel data structures employed by ADAMS are discussed, as is the client/server architecture. We list a number of sources of parallel speed-up found in typical ADAMS programs, and explain how these opportunities are exploited. Several potential future research projects related to this work are given. 1.

We investigate clustering techniques that are specifically tailored for object-oriented database systems. Unlike traditional database systems object-oriented data models incorporate the application behavior in the form of type-associated operations. This source of information is exploited for clustering decisions by statically determining the operations' access behavior applying dataflow analysis techniques. This process yields a set of weighted access paths. The statically extracted (syntactical) access patterns are then matched with the actual object net. Thereby the interobject reference chains that are likely being traversed in the database applications accumulate correspondingly high weights. The object net can then be viewed as a weighted graph whose nodes correspond to objects and whose edges are weighted inter-object references. We then employ a newly developed (greedy) heuristics for graph partitioning---which exhibits moderate complexity and, thus, is applicable t...

The name management capabilities currently provided by most existing persistent object systems (POSs) are rather limited. In particular, existing POSs tend to lack powerful and general mechanisms for forming, manipulating, controlling and reasoning about contexts. As a result, these POSs offer only weak or awkward support for large-scale data storage, multi-user computing, code reuse, interoperation of independently developed object stores and other similarly important classes of applications. As part of our work on improved name management for convergent computing systems, we have developed a framework for uniform treatment of the context and interface control facets of name management. In this paper we describe a realization of that framework, in the form of a shell-style user interface to a POS, that we are using to experiment both with the framework itself and with enhanced context control capabilities for POSs. 1 Introduction A convergent computing system is any system in which t...

This paper addresses the problem of clustering complex data on disk to minimize the number of I/O in data intensive applications. It describes the clustering strategies adopted in the O 2 system. As clustering depends on both structural aspects (composition hierarchy of the classes) and dynamic aspects (the methods associated with the classes) the paper details a cost model in order to evaluate the benefits of the clustering strategies. This model will permit to automatically derive new clustering strategies. To this end, a derivation algorithm which builds an optimal strategy in linear time is presented. Last, we give a short summary of past and future performance measurements activities. 1 Introduction In the past five years, object-oriented database systems have evolved to what is to become the major commercial database technology of the 1990's, [Ba88], [Ki88]. Whereas the relational model provided the user only with sets of tuples as data modelling primitives, object-oriented data...

A key factor for the outstanding success of database management

Clustering is one of the most effective means to enhance the performance of object base applications. Consequently, many proposals exist for algorithms computing good object placements depending on the application profile. However, in an effective object base reorganization tool the clustering algorithm is only one constituent. In this paper, we report on our object base reorganization tool that covers all stages of reorganizing the objects: the application profile is determined by a monitoring tool, the object placement is computed from the monitored access statistics utilizing a variety of clustering algorithms and, finally, the reorganization tool restructures the object base accordingly. The costs as well as the effectiveness of these tools is quantitatively evaluated on the basis of the OO1-benchmark. 1 Introduction Ever since the "early days" of database management systems, clustering has proven to be one of the most effective performance enhancement techniques. Therefore, many...

Due to the complexity of real-world applications, the number of databases and the volume of data have increased tremendously. Discovering qualitative and quantitative patterns from databases in such a distributed information-providing environment has been recognized as a challenging task. In response to such a demand, data mining and data warehousing techniques are emerging to extract the previously unknown and potentially useful knowledge to provide better decision support. This paper presents a mechanism called Markov Model Mediators (MMMs) to facilitate the understanding of the data warehouse schemas/views and the improvement of the query processing performance by analyzing and discovering the summarized knowledge at the database level. Simulation results show that the data mining process leads to a better federation of data warehouses and reduces the cost of query processing. To illustrate these benefits, our approach has been implemented and a simple example and several experime...

We present some results and raise some issues regarding benchmarks for evaluating garbage collection in persistent distributed stores. We believe that such issues are relevant due to increasing research activity in this domain. Therefore, defining widely accepted benchmarks to evaluate the performance and impact of garbage collection solutions is urgent. Currently, the performance of garbage collectors is measured using synthetic, often unportable and/or controversial benchmarks, making the relative evaluation of each solution almost impossible. In addition, there is a lack of knowledge on the behavior of real applications, e.g., amount of garbage, rate of death, etc. This paper outlines our experience on benchmarking our prototypes. We also report what we think is relevant to measure and which missing information we would like to obtain. 1 Introduction It has long been recognized that explicit storage management (e.g., malloc() and free()) increases code complexity and is highly erro...