Chinya V. Ravishankar

Spatial hash-joins

We examine how to apply the hash-join paradigm to spatial joins, and define a new framework for spatial hash-joins. Our spatial partition functions have two components: a set of bucket extents and an assignment function, which may map a data item into multiple buckets. Furthermore, the partition functions for the two input datasets may be different.We have designed and tested a spatial hash-join method based on this framework. The partition function for the inner dataset is initialized by sampling the dataset, and evolves as data are inserted. The partition function for the outer dataset is immutable, but may replicate a data item from the outer dataset into multiple buckets. The method mirrors relational hash-joins in other aspects. Our method needs no pre-computed indices. It is therefore applicable to a wide range of spatial joins.Our experiments show that our method outperforms current spatial join algorithms based on tree matching by a wide margin. Further, its performance is superior even when the tree-based methods have pre-computed indices. This makes the spatial hash-join method highly competitive both when the input datasets are dynamically generated and when the datasets have pre-computed indices.

Using name-based mappings to increase hit rates

Clusters of identical intermediate servers are often created to improve availability and robustness in many domains. The use of proxy servers for the World Wide Web (WWW) and of rendezvous points in multicast routing are two such situations. However, this approach can be inefficient if identical requests are received and processed by multiple servers. We present an analysis of this problem, and develop a method called the highest random weight (HRW) mapping that eliminates these difficulties. Given an object name and a set of servers, HRW maps a request to a server using the object name, rather than any a priori knowledge of server states. Since HRW always maps a given object name to the same server within a given cluster, it may be used locally at client sites to achieve consensus on object-server mappings. We present an analysis of HRW and validate it with simulation results showing that it gives faster service times than traditional request allocation schemes such as round-robin or least-loaded, and adapts well to changes in the set of servers. HRW is particularly applicable to domains in which there are a large number of requestable objects, there is a significant probability that a requested object will be requested again, and the CPU load due to any single object can be handled by a single server. HRW has now been adopted by the multicast routing protocols PIMv2 and CBTv2 as its mechanism for routers to identify rendezvous points/cores.

Spatial joins using seeded trees

Existing methods for spatial joins assume the existence of indices for the participating data sets. This assumption is not realistic for applications involving multiple map layer overlays or for queries involving non-spatial selections. In this paper, we explore a spatial join method that dynamically constructs index trees called seeded trees at join time. This methods uses knowledge of the data sets involved in the join process. Seeded trees are R-tree like structures, and are divided into the seed levels and the grown levels. The nodes in the seed levels are used to guide tree growth during tree construction. The seed levels can also be used to filter out some input data during construction, thereby reducing tree size. We develop a technique that uses intermediate linked lists during tree construction and significantly speeds up the tree construction process. The technique allows a large number of random disk accesses during tree construction to be replaced by smaller numbers of sequential accesses. Our performance studies show that spatial joins using seeded trees outperform those using other methods significantly in terms of disk I/O. The CPU penalties incurred are also lower except when seed-level filtering is used.

Distributed center-location algorithms

Recent multicast routing protocol proposals such as protocol independent multicast (PIM) and core-based trees (CBT) have been based on the notion of group-shared trees. Since construction of a minimal-cost tree spanning for all members of a group is difficult, they rely on center-based trees and distribute packets from all sources over a single shortest-path tree rooted at some center. PIM and CBT provisionally use administrative selection or simple heuristics for locating the center of a group but do not preclude the use of other methods that provide an ordered list of centers. Other previously proposed heuristics typically require knowledge of the complete network topology, a requirement which is not always practical for a distributed problem such as Internet routing. We investigate the problem of finding a good center in a distributed fashion, study various heuristics for automating center selection, and examine their applicability to real-world networks. We also propose several new algorithms which we feel to be more practical than existing methods. We present simulation results on hierarchical and nonhierarchical networks showing that of the methods potentially feasible in the Internet multicast backbone, ours offer the best results in terms of cost and delay, and they incur low overhead.

Monitoring and debugging distributed real-time programs

In this paper we describe the design and implementation of an integrated monitoring and debugging system for a distributed real‐time computer system. The monitor provides continuous, transparent monitoring capabilities throughout a real‐time systems lifecycle with bounded, minimal, predictable interference by using software support. The monitor is flexible enough to observe both high‐level events that are operating system‐ and application‐specific, as well as low‐level events such as shared variable references. We present a novel approach to monitoring shared variable references that provides transparent monitoring with low overhead. The monitor is designed to support tasks such as debugging realtime applications, aiding real‐time task scheduling, and measuring system performance. Since debugging distributed real‐time applications is particularly difficult, we describe how the monitor can be used to debug distributed and parallel applications by deterministic execution replay.

Efficient key establishment for group-based wireless sensor deployments

Establishing pairwise keys for each pair of neighboring sensors is the first concern in securing communication in sensor networks. This task is challenging because resources are limited. Several random key predistribution schemes have been proposed, but they are appropriate only when sensors are uniformly distributed with high density. These schemes also suffer from a dramatic degradation of security when the number of compromised sensors exceeds a threshold. In this paper, we present a group-based key predistribution scheme, GKE, which enables any pair of neighboring sensors to establish a unique pairwise key, regardless of sensor density or distribution. Since pairwise keys are unique, security in GKE degrades gracefully as the number of compromised nodes increases. In addition, GKE is very efficient since it requires only localized communication to establish pairwise keys, thus significantly reducing the communication overhead. Our security analysis and performance evaluation illustrate the superiority of GKE in terms of resilience, connectivity, communication overhead and memory requirement.

Block-oriented compression techniques for large statistical databases

Disk I/O has long been a performance bottleneck for very large databases. Database compression can be used to reduce disk I/O bandwidth requirements for large data transfers. The authors explore the compression of large statistical databases and propose techniques for organizing the compressed data such that standard database operations such as retrievals, inserts, deletes and modifications are supported. They examine the applicability and performance of three methods. Two of these are adaptions of existing methods, but the third, called tuple differential coding (TDC), is a new method that allows conventional access mechanisms to be used with the compressed data to provide efficient access. They demonstrate how the performance of queries that involve large data transfers can be improved with these database compression techniques.

A framework for pursuit evasion games in R n

We present a framework for solving pursuit evasion games in Rn for the case of N pursuers and a single evader. We give two algorithms that capture the evader in a number of steps linear in the original pursuer-evader distances. We also show how to generalize our results to a convex playing field with finitely many hyperplane boundaries that serve as obstacles.

Indexing Spatio-Temporal Trajectories with Efficient Polynomial Approximations

Complex queries on trajectory data are increasingly common in applications involving moving objects. MBR or grid-cell approximations on trajectories perform suboptimally since they do not capture the smoothness and lack of internal area of trajectories. We describe a parametric space indexing method for historical trajectory data, approximating a sequence of movement functions with single continuous polynomial. Our approach works well, yielding much finer approximation quality than MBRs. We present the PA-tree, a parametric index that uses this method, and show through extensive experiments that PA-trees have excellent performance for offline and online spatio-temporal range queries. Compared to MVR-trees, PA-trees are an order of magnitude faster to construct and incur I/O cost for spatio-temporal range queries lower by a factor of 2-4. SETI is faster than our method for index construction and timestamp queries, but incurs twice the I/O cost for time interval queries, which are much more expensive and are the bottleneck in online processing. Therefore, the PA-tree is an excellent choice for both offline and online processing of historical trajectories

On optimal processor allocation to support pipelined hash joins

In this paper, we develop algorithms to achieve optimal processor allocation for pipelined hash joins in a multiprocessor-based database system. A pipeline of hash joins is composed of several stages, each of which is associated with one join operation. The whole pipeline is executed in two phases: (1) the table-building phase, and (2) the tuple-probing phase. We focus on the problem of allocating processors to the stages of a pipeline to minimize the query execution time. We formulate the processor allocation problem as a two-phase mini-max optimization problem, and develop three optimal allocation schemes under three different constraints. The effectiveness of our problem formulation and solution is verified through a detailed tuple-by-tuple simulation of pipelined hash joins. Our solution scheme is general and applicable to any optimal resource allocation problem formulated as a two-phase mini-max problem.