Monir Abdullah

A2DLT: Divisible Load Balancing Model for Scheduling Communication-Intensive Grid Applications

Scheduling an application in data grid is significantly complex and very challenging because of its heterogeneous in nature of the grid system. Divisible Load Theory (DLT) is a powerful model for modelling data-intensive grid problem where both communication and computation loads are partitionable. This paper presents a new divisible load balancing model known as adaptive ADLT (A2DLT) for scheduling the communication intensive grid applications. This model reduces the maximum completion time (makespan) as compared to the ADLT and Constraint DLT (CDLT) models. Experimental results showed that the model can balance the load efficiently, especially when the communication-intensive applications are considered.

A novel economical duty cycle division multiplexing with electrical multiplexer and demultiplexer for optical communication systems

A duty cycle division multiplexing (DCDM) is proposed as an alternative multiplexing technique. It can be applied in either wired or wireless communication systems, although the focus in this paper is in optical fiber communications. The channel multiplexing and demultiplexing is performed electrically at single user bit rate which is very economic. In this paper, we examine 3 channels each operating at 10 Gb/s over a single optical carrier. The performance comparison is made against return-to-zero (RZ) transmitted pulses. Back-to-back receiver sensitivity and required optical signal-to-noise ratio (OSNR) of this system are examined and compared with 10 Gb/s RZ pulses. Effect of the chromatic dispersion is tested for DCDM channels and compared with 30 Gb/s RZ coding at the same transmission power. Also, the launched power of these two techniques is measured against system dispersion at BER of 10-9. The results show that, DCDM can support higher amount of chromatic dispersion than that RZ technique. In the end, receiver sensitivity and OSNR of 3-channel DCDM is tested at different bit rate of 2.5, 10, 25 and 40 Gb/s. A receiver sensitivity and OSNR of -16.8 dBm and 34.6 dB is required for the worst DCDM user when the system running at 3 times 40 Gb/s respectively.

Efficient parallel routing algorithms in optical multistage interconnection network

The crosstalk problem is introduced by an optical multistage interconnection network, which is caused by coupling two signals within a switching element. To avoid this crosstalk, a time domain approach is used, which is to partition the set of the connections into several subsets such that the connections in each subset can be established simultaneously in the network without crosstalk. Since we want to partition the messages to be sent to the network into several groups, we have to use a window method that is used for finding the conflicts among all the messages to be sent. In this paper, we proposed two parallel window method algorithms called unbalanced parallel window method (UPWM) and balanced parallel window method (BPWM) algorithms. The UPMW algorithm reduced the execution time approximately 79%, whereas the BPMW algorithm reduced 85% of the time compared to the sequential algorithm if we use seven processors.

Duty Cycle Division Multiplexing (DCDM); A New Electrical Multiplexing Technique for High Speed Optical Communication Systems

A new multiplexing technique based on duty cycle division is proposed, thus the name duty cycle division multiplexing (DCDM). DCDM can be applied in both electrical and optical domains, for wired and wireless systems. The new technique allows for more efficient use of time slots as well as the spectrum, taking advantage of both the conventional TDM and FDM. In this paper, three channels operating at the same speed of 10 Gbps per channel are multiplexed in the electrical domain. The performance comparison is made against 3times10 Gbps TDM, and the experimental simulation results show that the DCDM system can support higher bit rate than TDM and also, it is less sensitive to the chromatic dispersion effect.

Simulated annealing algorithm for scheduling Divisible Load in large scale data grids

In many data grid applications, data can be decomposed into multiple independent sub datasets and distributed for parallel execution and analysis. This property has been successfully exploited using divisible load theory (DLT). Many scheduling approaches have been studied but there is no optimal solution. This paper proposes a novel simulated annealing (SA) algorithm for scheduling divisible load in large scale data grids. SA algorithm is integrated with DLT model and compared with the previous approaches. Experimental results show that the proposed model obtains better solution in term of makespan.

Absolute Polar Duty Cycle Division Multiplexing (APDCDM); Technique for wireless communications

A new multiplexing and demultiplexing technique for wireless communications which is called Absolute Polar Duty Cycle Division Multiplexing (APDCDM) is presented in this paper. APDCDM can become an alternative multiplexing technique in wireless communications. The new technique allows for better error detection, correction, clock recovery and more efficient use of time slots as well as spectrum. The principle of the APDCDM technique has been discussed in this paper based on theoretical analysis as well as simulation studies. The performance comparison is made against time division multiplexing technique (TDM). The simulation has been set for wireless transmission, based on free space propagation model with adaptive white Gaussian noise (AWGN); QAM is used as modulation scheme to evaluate this technique against data rate and number of users. The simulation result correspond with the theoretical study show that APDCDM has better performance than TDM for supporting higher number of multiplexing users and bit rate.

Multi Slot Amplitude Coding Performance improvement with level spacing optimization

We reported an optimization method to improve Multi Slot Amplitude Coding (MSAC) optical fiber system performance. Simulation result clearly shows that optimum back to back BER performance is achieved when signal level spacing for upper level is higher than that of the lower level in order to compensate the signal dependent noise. In addition, we compare optimized system performance in terms of optical received power (ORP), optical signal to noise ratio (OSNR) and chromatic dispersion (CD) tolerance with conventional MSAC method. Improvement of 4.5 dB in ORP and OSNR are observed.

Probability of symbol error of Multi Slot Amplitude Coding (MSAC) technique in optical communication system

This paper reports a novel method on the probability of symbol error estimation for high speed Multi Slot Amplitude Coding (MSAC) technique in optically amplified communication system. The system setup is based on simple intensity modulation and direct detection scheme which is cost effective. In this proposed method, possibility of symbol error is estimated based on individual probability density function of particular signal level and slot of MSAC symbol. This proposed method is validated by the symbol-to-symbol error technique.

New Optimal Load Allocation for Scheduling Divisible Data Grid Applications

In many data grid applications, data can be decomposed into multiple independent sub-datasets and distributed for parallel execution and analysis. This property has been successfully employed by using Divisible Load Theory (DLT), which has been proved as a powerful tool for modeling divisible load problems in data-intensive grid. There are some scheduling models have been studied but no optimal solution has been reached due to the heterogeneity of the grids. This paper proposes a new model called Iterative DLT (IDLT) for scheduling divisible data grid applications. Recursive numerical closed form solutions are derived to find the optimal workload assigned to the processing nodes. Experimental results show that the proposed IDLT model obtains better solution than other models (almost optimal) in terms of makespan .

A new load balancing scheduling model in data grid application

Scheduling an application in data grid is significantly complex and very challenging because of its heterogeneous in nature of the grid system. Thus, Divisible Load Theory (DLT) is a powerful model for modeling data intensive grid problem where both communication and computation load is partitionable. Previously, Task Data Present (TDP) model was proposed based on DLT model. This paper presents an Adaptive TDP (ATDP) model to reduce the makespan. New equations for calculating the load allocation are derived. Experimental results showed that the proposed model can balance the load efficiently.