Hardware trends have produced an increasing disparity between processor speeds and memory access times. While a variety of techniques for tolerating or reducing memory latency have been proposed, these are rarely successful for pointer-manipulating programs. This paper explores a complementary approach that attacks the source (poor reference locality) of the problem rather than its manifestation (memory latency). It demonstrates that careful data organization and layout provides an essential mechanism to improve the cache locality of pointer-manipulating programs and consequently, their performance. It explores two placement technique-lustering and colorinet improve cache performance by increasing a pointer structure’s spatial and temporal locality, and by reducing cache-conflicts. To reduce the cost of applying these techniques, this paper discusses two strategies-cache-conscious reorganization and cacheconscious allocation--and describes two semi-automatic toolsccmorph and ccmalloc-that use these strategies to produce cache-conscious pointer structure layouts. ccmorph is a transparent tree reorganizer that utilizes topology information to cluster and color the structure. ccmalloc is a cache-conscious heap allocator that attempts to co-locate contemporaneously accessed data elements in the same physical cache block. Our evaluations, with microbenchmarks, several small benchmarks, and a couple of large real-world applications, demonstrate that the cache-conscious structure layouts produced by ccmorph and ccmalloc offer large performance benefit-n most cases, significantly outperforming state-of-the-art prefetching.

Thor is an object-oriented database system designed for use in a heterogeneous distributed environment. It provides highly-reliable and highly-available persistent storage for objects, and supports safe sharing of these objects by applications written in different programming languages. Safe heterogeneous sharing of long-lived objects requires encapsulation: the system must guarantee that applications interact with objects only by invoking methods. Although safety concerns are important, most object-oriented databases forgo safety to avoid paying the associated performance costs. This paper gives an overview of Thor's design and implementation. We focus on two areas that set Thor apart from other objectoriented databases. First, we discuss safe sharing and techniques for ensuring it; we also discuss ways of improving application performance without sacrificing safety. Second, we describe our approach to cache management at client machines, including a novel adaptive prefetching strat...

Object-oriented databases store many small objects on disks. Disks perform poorly when reading and writing individual small objects. This thesis presents a new storage management architecture that substantially improves disk performance of a distributed object-oriented database system. The storage architecture is built around a large modified object buffer (MOB) that is stored in primary memory. The MOB provides volatile storage for modified objects. Modified objects are placed in the MOB instead of being immediately written out to disk. Modifications are written to disk lazily as the MOB fills up and space is required for new modifications. The MOB improves performance because even if an object is modified many times in a short period of time, the object has to be written out to disk only once. Furthermore, by the time an object modification has to be flushed from the MOB, many modifications to other objects on the same page may have accumulated. All of these modifications can be writ...

Many client-server object-oriented database systems (OODBs) run applications at clients and perform all accesses on cached copies of database objects. Moving both data and computation to the clients can improve response time, throughput, and scalability. For applications with good locality of reference, retaining cached state across transaction boundaries can result in further performance and scaling benefits. This thesis examines the question of what concurrency control scheme is best able to realize these potential benefits. It describes a new optimistic concurrency control scheme called AOCC (Adaptive Optimistic Concurrency Control) and compares its performance with that of ACBL (Adaptive-Granularity Callback Locking), the scheme shown to have the best performance in previous studies. Like all optimistic schemes, AOCC synchronizes transactions at the commit point, aborting transactions when synchronization fails; ACBL, like other locking schemes, synchronizes transactions while they execute. Earlier

Current commercial databases allow application programmers to trade off consistency for performance. However, existing definitions of weak consistency levels are either imprecise or they disallow efficient implementation techniques such as optimism. Ruling out these techniques is especially unfortunate because commercial databases support optimistic mechanisms. Furthermore, optimism is likely to be the implementation technique of choice in the geographically distributed and mobile systems of the future. This thesis presents the first implementation-independent specifications of existing ANSI isolation levels and a number of levels that are widely used in commercial systems, e.g., Cursor Stability, Snapshot Isolation. It also specifies a variety of guarantees for predicate-based operations in an implementation-independent manner. Two new levels are defined that provide useful consistency guarantees to application writers; one is the weakest level that ensures consistent reads, while the other captures some useful consistency properties provided by pessimistic implementations. We

Parameter passing is one of the main problems in distributed object-oriented applications. The two simplest solutions - passing objects by global reference and passing complete copies of the object graphs - both have significant drawbacks. Instead, an intermediate amount of copying is often best. The problem is twofold: (i) compilers and operating systems can't automatically make the best decision on how much to copy; (ii) there hasn't been a good mechanism for the programmer to express the intermediate amount of copying that should be done. The contribution of this paper is to present a new mechanism that allows programmers to express in a natural and succinct way how much of the object graph should be copied to the remote context. The specifications are done using a very simple declarative meta-language -- GOOP -- that has graphs of classes as domain. Keywords: distribution, remote invocation, pass-by-copy, open implementation, adaptiveness. Supported by the Portuguese Fo...

We classify clustering algorithms into sequence-based techniques---which transform the object net into a linear sequence---and partition-based clustering algorithms. Tsangaris and Naughton [TN91, TN92] have shown that the partition-based techniques are superior. However, their work is based on a single partitioning algorithm, the Kernighan and Lin heuristics, which is not applicable to realistically large object bases because of its high running-time complexity. The contribution of this paper is two-fold: (1) we devise a new class of greedy object graph partitioning algorithms (GGP) whose running-time complexity is moderate while still yielding good quality results. For large object graphs GGP is the best known heuristics with an acceptable running-time. (2) We carry out an extensive quantitative analysis of all well-known partitioning algorithms for clustering object graphs. Our analysis yields that no one algorithm performs superior for all object net characteristics. Therefore, we d...

This report makes a case for more expressive interfaces between operating systems (OSes) and storage devices. In today's systems, the storage interface consists mainly of simple read and wri t e commands; as a result, OSes operate with little understanding of device-specific characteristics and devices operate with little understanding of system priorities. More expressive interfaces, together with extended versions of today's OS and firmware specializations, would allow the two to cooperate to achieve performance and functionality that neither can achieve alone.

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...

There are a variety of main-memory access structures, such as segment trees, and quad trees, whose properties, such as good worst-case behaviour, make them attractive for database applicdions. Unfortunately, the structures are typically ‘long and skinny’, whereas disk data structuies must be ‘short-and-fat (that is, have a high fanout and low height) in order to minimize I/O. We consider how to cluster the nodes (that is, map the nodes to disk pages) of main-memory access structures such that although a path may traverse many nodes, it only tra-verses a few disk pages. The number of disk pages traversed in a path is called the exter-nal path length. We address several versions of the clustering problem. We present a clus-tering algorithm for tree structures that gener-ates optimal worst-case external path length mappings; we also show how to make it dy-namic, to support updates. We extend the al-gorithm to generate mappings that minimize the average weighted external path lengths. We also show that some other clustering prob-lems, such as finding optimal external path lengths for DAG structures and minimizing