1. Introduction

2. Background: Metrics for detecting influential nodes in large biological networks

3. Proposed parallel algorithm for detecting influential nodes in large biological networks

4. Results

5. Discussion

6. Conclusion

Acknowledgments

References

Abstract

In biological networks, some nodes are more influential than others. The most influential nodes are those whose elimination induces a network collapse, and detecting these nodes is crucial in many circumstances. However, this is a difficult task when the size of the biological networks is large. In this paper, we have designed and implemented an efficient parallel algorithm for detecting influential nodes for large biological networks by exploiting a Graphics Processing Unit (GPU). The essential concept behind the proposed parallel algorithm is that several computationally expensive procedures in detecting influential nodes are redesigned and transformed into quite efficient GPUaccelerated primitives such as parallel sort, scan, and reduction. Four local metrics, including the Degree Centrality (DC), Companion Behavior (CB), Clustering Coefficient (CC), and H-Index, are used to measure the nodal influence. To evaluate the efficiency of the proposed parallel algorithm, five large real biological networks are employed in the experiments. The experimental results show that (1) the proposed parallel algorithm can achieve speedups of approximately 48∼94 over the corresponding serial algorithm; (2) compared to a baseline parallel algorithm developed on a multi-core CPU, the proposed parallel algorithm yields speedups of 5∼9 for DC and H-Index, while it is slightly slower for CB and CC due to the uneven degree distribution; and (3) when using DC and H-Index, the proposed parallel algorithm is capable of detecting the influential nodes in a large biological network consisting of 150 million edges in less than 3 s.

Introduction

In recent years, complex network analysis has received increasing attention. Many real complex systems can be abstractly regarded as complex networks for presenting the complexities of real systems [1], such as social networks, technological networks, information networks, and biological networks. Various methods have been proposed for mining information from complex networks. In particular, detecting the influential nodes in complex networks is a topic of interest drawing much attention in this research field [2–4].

Detecting influential nodes in complex networks can be exploited to mine the features and functions of these networks [2]. Much research has been conducted to rank and identify influential nodes in the aforementioned four network categories. In the detection of influential nodes in various networks, the first critical issue is to select or define the metrics for measuring the influence of each node; the second is to employ or develop specific algorithms to effectively and efficiently determine and rank the influential nodes.

Many metrics have been proposed for detecting the influential nodes in complex networks. These metrics can be roughly divided into two categories [4]: (1) local metrics that are calculated based on the local structures of networks and (2) global metrics that are calculated based on the global structures of networks. For the first category, the most commonly used metrics include the degree centrality, clustering coefficient, and H-Index [5]. For the second category, the most commonly used metrics are the betweenness centrality [6], closeness centrality, PageRank, k-core [7] computed using the k-shell decomposition [8], and the bidirectional k-core (B-core) [9].

In the quest for exascale computing, energy-efficiency is a fundamental goal in highperformance computing systems, typically achieved via dynamic voltage and frequency scaling (DVFS). However, this type of mechanism relies on having accurate methods of predicting the performance and power/energy consumption of such systems. Unlike previous works in the literature, this research focuses on creating novel GPU predictive models that do not require run-time information from the applications. The proposed models, implemented using recurrent neural networks, take into account the sequence of GPU assembly instructions (PTX) and can accurately predict changes in the execution time, power and energy consumption of applications when the frequencies of different GPU domains (core and memory) are scaled. Validated with 24 applications on GPUs from different NVIDIA microarchitectures (Turing, Volta, Pascal and Maxwell), the proposed models attain a significant accuracy. Particularly, the obtained power consumption scaling model provides an average error rate of 7.9% (Tesla T4), 6.7% (Titan V), 5.9% (Titan Xp) and 5.4% (GTX Titan X), which is comparable to state-of-the-art run-time counter-based models. When using the models to select the minimum-energy frequency configuration, significant energy savings can be attained: 8.0% (Tesla T4), 6.0% (Titan V), 29.0% (Titan Xp) and 11.5% (GTX Titan X).

Introduction

Over the past decade, the high-performance computing (HPC) area has observed a noticeable upsurge in the utilization of accelerators, more specifically graphics processing units (GPUs). The energy efficiency of these devices can have a large impact on the total cost of large-scale computer clusters. As an example, the Summit supercomputer (number one system of June’2019 Top500 list [1]), uses a total of 27 648 NVIDIA Volta GPUs to achieve a peak performance of almost 200 petaflops. For that, it requires a power supply of 13 million watts, which corresponds to an estimated cost of 17 million dollars per year (on power supply alone) [2]. The magnitude of such values highlights the importance of effective mechanisms to maximize the energy efficiency of these systems, as a mere 5% decrease in the energy consumption could generate savings of around 1 million dollars. One example of such mechanisms is the dynamic voltage and frequency scaling (DVFS), which allows placing devices into lower performance/power states. When carefully applied to match the needs of the executing applications, DVFS can lead to significant power and energy savings, sometimes with minimum impact on performance [3], [4]. A recent study showed that using DVFS techniques in GPUs executing deep neural networks applications can provide energy savings up to 23% during training and 26% during inference phases [5].

1- Introduction

2- Related work

3- TFL strategy

4- HIL Strategy

5- Multiple GPUs and cost model

6- Performance evaluation

7- Summary

References

Abstract

Frequent itemset mining is widely used as a fundamental data mining technique. However, as the data size increases, the relatively slow performances of the existing methods hinder its applicability. Although many sequential frequent itemset mining methods have been proposed, there is a clear limit to the performance that can be achieved using a single thread. To overcome this limitation, various parallel methods using multi-core CPU, multiple machine, or many-core graphic processing unit (GPU) approaches have been proposed. However, these methods still have drawbacks, including relatively slow performance, data size limitations, and poor scalability due to workload skewness. In this paper, we propose a fast GPU-based frequent itemset mining method called GMiner for large-scale data. GMiner achieves very fast performance by fully exploiting the computational power of GPUs and is suitable for large-scale data. The method performs mining tasks in a counterintuitive way: it mines the patterns from the first level of the enumeration tree rather than storing and utilizing the patterns at the intermediate levels of the tree. This approach is quite effective in terms of both performance and memory use in the GPU architecture. In addition, GMiner solves the workload skewness problem from which the existing parallel methods suffer; as a result, its performance increases almost linearly as the number of GPUs increases. Through extensive experiments, we demonstrate that GMiner significantly outperforms other representative sequential and parallel methods in most cases, by orders of magnitude on the tested benchmarks.

Introduction

As a fundamental data mining technique, frequent itemset mining is widely used in a wide range of disciplines such as market basket analysis, web usage mining, social network analysis, intrusion detection, bioinformatics, and recommendation systems. However, the deluge of data generated by automated systems for diagnostic or analysis purposes makes it difficult or even impossible to apply mining techniques in many realworld applications. The existing methods often fail to find frequent itemsets in such big data within a reasonable amount of time. Thus, in terms of computational time, itemset mining is still a challenging problem that has not yet been completely solved. Many sequential frequent itemset mining methods such as Apriori [2], Eclat [35], FP-Growth [14], and LCM [30] use a single CPU thread. However, these singlethreaded applications all have a fundamental mining performance limit because CPU clock speed is generally no longer increasing. To overcome the single-thread performance limit, multiple parallel frequent itemset mining methods have been proposed. These methods can be categorized into three main groups: (1) (CPU-based) multithreaded methods, (2) distributed methods, and (3) graphic processing unit (GPU)- based methods. We omit the term ”multi-thread” from the GPU-based methods because they are obviously multi-threaded. The first group focuses on accelerating the performance of the single-threaded methods by exploiting multi-core CPUs [20, 24, 26, 27, 29], while the second group tries to accelerate the performance by exploiting multiple machines [12, 17, 19]. Details about these methods are available in recent survey studies.