Aihua Liang

Arithmetic Computation Using Self-Assembly of DNA Tiles: Modular-Square over Finite Field GF(2n)

Tile assembly model is a highly distributed parallel model of DNA computing. This paper proposes how the tile assembly process could be used for computing the modular square, an operation combining square and reduction, over finite field GF(2n). In this molecular computing system, reduction is executed after the completion of square. The time complexity of this molecular computing system is ⊙(n) and the space complexity is ⊙(n2). This system requires 75 types of computation tiles and 9 types of boundary tiles.

Energy Aware Scheduling for Precedence Constrained Parallel Tasks in a Power-Scalable Cluster

Improving the energy efficiency of high performance clusters has become important research issue. We proposed a new algorithm that reduces energy consumption of precedence constrained parallel tasks in power-scalable clusters. To reduce energy consumption without increasing the schedule length, our algorithm reclaims both static and dynamic slack time and employs different frequency adjusting techniques in different slack time. The optimal frequency is obtained through analyzing the precedence constraints of parallel tasks. We conducted experiments to compare the proposed algorithm with two other existing algorithms. Simulation results show that the proposed algorithm can get better energy efficiency without increasing the make span.

Adaptive workload driven dynamic power management for high performance computing clusters

With the scale expansion of high performance computer systems, efficient power management has developed into an important issue. To strive to balance power consumption and performance, this paper proposes an adaptive workload-driven dynamic power management policy for homogeneous clusters, which dynamically adjusts the power mode of computing nodes according to workload variation. The proposed policy combines the pre-wakeup method and the feedback mechanism to reduce performance degradation due to the wakeup delay. The experimental results demonstrate that, as compared with two existing timeout policies, adaptive workload-driven dynamic power management effectively reduced the performance loss with a slight increase in power consumption.

Workload-aware Power Management of Cluster Systems

Along with the increased awareness of energy cost, power management becomes a big issue for clusters. In this study, we investigate the workload aware power management techniques for cluster systems and propose a new power management policy. The considered power management techniques are dynamic workload consolidation and usage of dynamic power range enabled by low power states on servers. The power management policy has been implemented on OpenPBS with Maui, which reduces about ten percent power consumption and optimizes workload distribution to lower the impact on performance.

A Parallel Workload Model and its Implications for Maui Scheduling Policies

We develop a workload model based on the logs of real workloads of the Linux cluster Atlas and a small IBM Blue Gene/L cluster at Lawrence Livermore National Laboratory (LLNL), and the 184-node IBM eServer pSeries 655/690 at the San Diego Supercomputer Center (SDSC). This model gives us insight into the performance of scheduling jobs on space-sharing parallel computers, provided by Maui. We find out that backfill queuing policies improve a lot of the system performance, and without reservation, the allocation policies do not have an obvious distance between each other.

Fast Parallel Garner Algorithm for Chinese Remainder Theorem

This paper presents a fast parallel garner algorithm for Chinese remainder theorem. The variables in garner algorithm are divided into public parameters that are constants for fixed module and private parameters that represent random input integers. We design the parallel garner algorithm by analyzing the data dependencies of these arithmetic operations for computing public variables and private variables. Time complexities and speedup ratios of the parallel algorithm and the sequential algorithm are calculated to make the quantitative comparison based on our previous work about some fundamental parallel algorithms. The performance evaluation shows high efficiency of the proposed parallel algorithm compared to the sequential one.

An integrated resource allocation algorithm for CPU intensive and I/O intensive jobs in multi-core cluster

In cluster jobs have different characteristics, such as CPU intensive and I/O intensive. Many resource allocation strategies deal with CPU intensive jobs or I/O intensive jobs. However, few algorithms can deal with both of them. In this paper we propose an integrated resource allocation algorithm, which considers job characteristics. The proposed algorithm is based on an overhead model. The model can integrate the different job characteristics. Simulation results indicate that the proposed resource algorithm can effectively achieve improvement in average execution time compared with FIRSTAVAILABLE. Moreover, proposed algorithm can promote system load balancing to some extent.