Browsing by Author "Lu Hongjun"

Now showing 1 - 5 of 5
  • No Thumbnail Available
    Item
    Double-sided Composition of Relations
    (1992-03-01T00:00:00Z) Wang Ke; Zhang Weining; Lu Hongjun
    Abstract not available.
  • No Thumbnail Available
    Item
    Dynamic Load Balanced Join Processing
    (1993-10-01T00:00:00Z) Tan K L; Lu Hongjun
    In this paper, we revisit the problem of processing joins in shared-nothing systems. In such systems, a join query is usually split into a set of tasks that are allocated to the nodes in the system where the tasks could be executed concurrently and independently. While parallel processing could greatly reduce the completion time of a join operation, the system performance may degrade because of load imbalance across the nodes caused by data skewness in the relations. Most of the previous studies addressed this issue by static load balancing. We proposed here two dynamic load balancing join strategies and compared their performance with two static algorithms. The result of our study shows that dynamically balancing the join load is not only feasible but also provides good system performance.
  • No Thumbnail Available
    Item
    Extensible Buffer Management of Indexes
    (1992-03-01T00:00:00Z) Chan C Y; Ooi B C; Lu Hongjun
    Abstract not available.
  • No Thumbnail Available
    Item
    Incorporating Pipelining into Query Optimizer
    (1993-04-01T00:00:00Z) Tan K L; Lu Hongjun
    Abstract not available.
  • No Thumbnail Available
    Item
    On Spatially Partitioned Temporal Join
    (1994-01-01T00:00:00Z) Lu Hongjun; Ooi Beng Chin; Tan Kian Lee
    Temporal relations are relations in temporal database which have one or more attributes whose values vary along the time. The most common representation of time adopted in such databases are the time intervals during which those attributes have specified values. Time join in a temporal database matches tuples from two temporal relations whose time intervals overlap. It is an important but very expensive operation because the volume of temporal data is usually large and the join is a non-equijoin operation in nature. This paper presents a partition-based time join method and its supporting storage structure. Under the proposed method, time intervals in a temporal relations are mapped to points in a two dimensional space. These points are stored in clustered buckets whose addresses can be easily computed with the help of a semi-dynamic directory. When a time join of two relations is to be performed, a bucket in one relation only need to be compared with a determinable set of buckets of the other relation whose addresses can also be efficiently computed. As such, the proposed algorithm outperforms both the nested-loops and sort-merge based time join algorithms. The time-space mapping mechanism, the storage structure, and the join algorithms are discussed. Some results of our preliminary performance study are also provided to show the efficiency of the proposed method.