Abstract: The K computer, released on September 29, 2012, is a large-scale parallel supercomputer system consisting of 82,944 compute nodes. We have been able to resolve a significant number of operation issues since its release. Some system software components have been fixed and improved to obtain higher stability and utilization. We achieved 94% service availability because of a low hardware failure rate and approximately 80% node utilization by careful adjustment of operation parameters. We found that the K computer is an extremely stable and high utilization system.

Abstract: Quantum mechanically calculated positron-electron annihilation gamma-ray spectra, C1s binding energy spectra and NMR spectra are employed to study the electronic structures of hexane and its isomers, which is assisted using graph theory. Our recent positron-electron annihilation gamma-ray spectral study of n-hexane in gas phase and core ionization (IPs) spectral studies for small alkanes and their isomers, have paved the path for the present correlation study where quantum mechanics is combined with graph theory, C1s ionization spectroscopy and nuclear magnetic resonance (NMR), to further understand the electronic structure and topology for the hexane isomers. The low-energy plane wave positron (LEPWP) model indicated that the positrophilic electrons of a molecule are dominated by the electrons in the lowest occupied valence orbital (LOVO). The most recent results using NOMO indicated that the electronic wave functions dominate the electron-positron wave functions for molecular systems. In addition to quantum mechanics, chemical graphs are also studied and are presented in the present study.

Abstract: In this paper, we are concerned with the problem of flooding undirected weighted graphs under ceiling constraints. We provide a new algorithm based on a hierarchical structure called {\em dendrogram}, which offers the significant advantage that it can be used for multiple flooding with various scenarios of the ceiling values. In addition, when exploring the graph through its dendrogram structure in order to calculate the flooding levels, independent sub-dendrograms are generated, thus offering a natural way for parallel processing. We provide an efficient implementation of our algorithm through suitable data structures and optimal organisation of the computations. Experimental results show that our algorithm outperforms well established classical algorithms, and reveal that the cost of building the dendrogram highly predominates over the total running time, thus validating both the efficiency and the hallmark of our method. Moreover, we exploit the potential parallelism exposed by the flooding procedure to design a multi-thread implementation. As the underlying parallelism is created on the fly, we use a queue to store the list of the sub-dendrograms to be explored, and then use a dynamic round-robin scheduling to assign them to the participating threads. This yields a load balanced and scalable process as shown by additional benchmark results. Our program runs in few seconds on an ordinary computer to flood graphs with more that $20$ millions of nodes.

Abstract: The demands of improving energy efficiency for high performance scientific applications arise crucially nowadays. Software-controlled hardware solutions directed by Dynamic Voltage and Frequency Scaling (DVFS) have shown their effectiveness extensively. Although DVFS is beneficial to green computing, introducing DVFS itself can incur non-negligible overhead, if there exist a large number of frequency switches issued by DVFS. In this paper, we propose a strategy to achieve the optimal energy savings for distributed matrix multiplication via algorithmically trading more computation and communication at a time adaptively with user-specified memory costs for less DVFS switches, which saves 7.5% more energy on average than a classic strategy. Moreover, we leverage a high performance communication scheme for fully exploiting network bandwidth via pipeline broadcast. Overall, the integrated approach achieves substantial energy savings (up to 51.4%) and performance gain (28.6% on average) compared to ScaLAPACK pdgemm() on a cluster with an Ethernet switch, and outperforms ScaLAPACK and DPLASMA pdgemm() respectively by 33.3% and 32.7% on average on a cluster with an Infiniband switch.

Abstract: An important challenge in the design of databases that support multi-tenant applications is to provide a platform to manage large volumes of data collected from different businesses, social media networks, emails, news, online texts, documents, and other data sources. To overcome this challenge we proposed in our previous work a multi-tenant database schema called Elastic Extension Tables (EET) that combines multi-tenant relational tables and virtual relational tables in a single database schema. Using this approach, the tenants’ tables can be extended to support the requirements of individual tenants. In this paper, we discuss the potentials of using EET multi-tenant database schema, and show how it can be used for managing physical and virtual relational data. We perform several experiments to measure the feasibility and effectiveness of EET by comparing it with a commercially available multi-tenant schema mapping technique used by SalesForce.com. We report significant performance improvements obtained using EET when compared to Universal Table Schema Mapping (UTSM), making the EET schema a good candidate for the management of multi-tenant data in Software as a Service (SaaS) and Big Data applications.