Siu-Cheung Chau

Data partition and parallel evaluation of Datalog programs

Parallel bottom-up evaluation provides an alternative for the efficient evaluation of logic programs. Existing parallel evaluation strategies are neither effective nor efficient in determining the data to be transmitted among processors. In this paper, we propose re different strategy, for general Datalog programs, that is based on the partitioning of data rather than that of rule instantiations. The partition and processing schemes defined in this paper are more general than those in existing strategies. A parallel evaluation algorithm is given based on the semi-naive bottom-up evaluation. A notion of potential usefulness is recognized as a data transmission criterion to reduce, both effectively and efficiently, the amount of data transmitted. Heuristics and algorithms are proposed for designing the partition and processing schemes for a given program. Results from an experiment show that the strategy proposed in this paper has many promising features. >

Cyclic-cubes: a new family of interconnection networks of even fixed-degrees

We introduce a new family of interconnection networks that are Cayley graphs with fixed degrees of any even number greater than or equal to four. We call the proposed graphs cyclic-cubes because contracting some cycles in such a graph results in a generalized hypercube. These Cayley graphs have optimal fault tolerance and logarithmic diameters. For comparable number of nodes, a cyclic-cube can have a diameter smaller than previously known fixed-degree networks. The proposed graphs can adopt an optimum routing algorithm known for one of its subfamilies of Cayley graphs. We also show that a graph in the new family has a Hamiltonian cycle and, hence, there is an embedding of a ring. Embedding of meshes and hypercubes are also discussed.

Shape representation and description using the Hilbert curve

In this paper, a novel linear-time approach to shape representation and description is presented. The object shape is captured by scanning the object image using a space-filling curve (SFC). The resulting vector is smoothed, using wavelet approximation, and sampled. In addition, the concept of key feature points (KFPs) is introduced to utilize a priori information about the classification of the images in the database in optimizing the representation of the objects within each class. The proposed technique achieves a recognition rate of 88.3% on the MPEG-7 core experiment part B. On the Kimia-99 and Kimia-216 datasets, a precision average of 95.6% is attained. Retrieval rates of 94.2% and 95.6% are achieved on the gray-scale and binary versions of the ETH-80 dataset, respectively.

Routing and scheduling for a novel optical multistage interconnection network

Multistage Interconnection Networks (MINs) are popular in computing and communication applications. Recently, there have also been significant advancesin electro-optic switches that have made Optical MINs (OMINs) a good choice for the high channel bandwidth and low communication latency of high performance computing and communication applications. However, OMINs introduce crosstalkwhich results fromcoupling two signals within oneswitching element. Under the constraint of avoiding crosstalk, what we are interested inis how to realize a permutation that requires theminimum number of passes. This routing problem is an NP-hard one, and manyheuristic algorithms have been devised to find a solution. In [9], Chau and Xiao have proposed an algorithm, called the Remove Last Pass (RLP) algorithm, to avoid crosstalk and route the traffic in an OMIN more efficiently. In this paper, we focus on the routing and scheduling of a novel OMIN, the base-2 MIN, propoundedby Chau and Fu in [8]. Our experiments provethat any permutation can be realized in no more than threepasses in the base-2 OMIN by using the RLP algorithm, when the network has no more than 512 nodes. The base-2 OMIN requires only n(logn+1) switching elements (SEs)for an n×n network, compared to the crossbar, which requires O(n2) SEs for an n×n network. Therefore,the base-2 MIN should be a good candidate for communication subsystems in a parallel computing environment.

An Efficient Shape Representation and Description Technique

In this paper, we present a novel approach to shape representation and description based on the combination of the Hilbert space filling curve and wavelet analysis. Our objective is to capitalize on the localization-preserving nature of the Hilbert space filling curve and the approximation power of the wavelet transform. The object image is scanned using the Hilbert curve and the resulting vector is smoothed using the wavelet transform and sampled. The technique is O(N) for both representation and comparison. We present some experimental results on the MPEG-7 dataset, Kimia-99 dataset, ETH-80 dataset, and a logo dataset.

A Gracefully Degradable Declustered RAID Architecture

A new layout method Prime-groups is proposed to evenly distribute parity groups for declustered RAID. Prime-groups satisfies most of the layout goals for a good declustered RAID layout. For the goals that are not satisfied, it is near optimal. A new layout goal maximal write and reconstruction parallelism is also proposed. If a layout satisfies the new goal, all the surviving disks can be read in parallel and can be rewritten in parallel during reconstruction and reconfiguration. Prime-groups satisfies the new goal when the write request begins in the first disk of the array. It is also near optimal in term of declustering ratio when p is a prime. Prime-groups can compliment the layouts proposed by Alvarez et al. [2] for the criteria to obtain a good layout is quite different between Prime-groups and those proposed by Alvarez et al. [2].

Data partition: a practical parallel evaluation of datalog programs

The authors analyse previous parallel evaluation strategies based on the partition of rule instantiations. They present a parallel evaluation strategy for general Datalog programs that is based on the partition of data and is more practical. A key issue is to determine a criterion of data transmission that reduces the amount of data transmitted and is tested efficiently. A notion of potential usefulness is given as such a criterion. The problem of designing appropriate partition schemes and processing schemes is addressed. Heuristics and algorithms are proposed for making decisions in the design process.<<ETX>>

A Fault-Tolerant Binary Tree Architecture

A new fault-tolerant binary tree architecture is proposed. The approach, employing redundant processors, is suitable for use in long-life unmaintained applications. In our proposed scheme, processors connected as n level full binary tree, are backed up by k spare processors so that the fault-tolerant binary tree can tolerate any set of k faults. This new scheme is compared to previously proposed fault-tolerant binary tree architectures. In particular, we show that the new scheme can achieve the same level of reliability as other proposed scheme while using significantly fewer spares.

Load Balancing between Heterogeneous Computing Clusters

More and more organizations have computing clusters located in different places. The distance between the computing clusters can be quite far away and the computing clusters can also differ in size. The load in one cluster may be very high while the other clusters may have nothing running on the system. A higher throughput can be achieved if load balancing is added. In this paper, we propose a simple and efficient load balancing method to balance loads for heterogeneous computing clusters of various sizes that are quite far away from each other. We assume that the computer clusters are connected in the form of a hypercube. We then use the load balancing method for hypercubes to balance the load between heterogeneous computing clusters. After applying the propose method, the maximum load difference between any two computing clusters can be at most one more than the optimal solution.

A gracefully degradable declustered RAID architecture with near optimal maximal read and write parallelism

A new layout method, Prime-groups, is proposed to evenly distribute parity groups for declustered RAID. Prime-groups satisfies most of the layout goals for a good declustered RAID layout. For the goals that are not satisfied, it is near optimal. A new layout goal maximal write and reconstruction parallelism is also proposed. If a layout satisfies the new goal, all the surviving disks can be read in parallel and can be rewritten in parallel during reconstruction and reconfiguration. Prime-groups satisfies the new goal when the write request begins in the first disk of the array. It is also near optimal in term of declustering ratio when p is a prime. Prime-groups can compliment the layouts proposed by G.A. Alvarez et. al. (1996), as the criteria to obtain a good layout is quite different between Prime-groups and those proposed by Alvarez et. al.