Search

Storage class memories (SCMs) realize fast access latency. However, SCMs have a trade-off between reliability and latency in case of correcting errors. To achieve high reliability and high performance, double asymmetric-latency SCM SSD has been proposed. Previously proposed SSD consists of fast-write/slow-read SCM (W-SCM) and fast-read/slow-write SCM (R-SCM). W-SCM and R-SCM are realized by changing strength of ECC and number of Set/Reset verify cycles by SSD controller. Previously proposed SSD with asymmetric read/write latency requires frequently written (write-hot) data to be allocated in W-SCM and frequently read (read-hot) data to be allocated in R-SCM to achieve high performance [1]. Conventional data allocation algorithm for previously proposed SSD [1] triggers frequently data movement and degrades SSD performance. Therefore, this paper proposes data allocation algorithm that restricts data movement and stores write-hot, read-hot and cold data separately in W-SCM and R-SCM based on access frequency. As a result, SSD performance is improved by 14.1% for write-hot workload and 67.7% for read-hot workload compared to conventional algorithm.

International audience; Because of the evolution of compute units, memory hetero-geneity is becoming popular in HPC systems. But dealing with such various memory levels often requires different approaches and interfaces. For this purpose, OpenMP 5.0 defines memory-management constructs to offer application developers the ability to tackle the issue of exploiting multiple memory spaces in a portable way. This paper proposes an overview of memory-management from applications to runtimes. Thus, we describe a convenient way to tune an application to include memory management constructs. We also detail a methodology to integrate them into an OpenMP runtime supporting multiple memory types (DDR, MC-DRAM and NVDIMM). We implement our design into the MPC framework , while presenting some results on a realistic benchmark.

Cloud computing provides a flexible and cost-effective solution to big data applications. Data privacy, however, is a main concern. To avoid information leakage to a cloud service provider, a user may store portions of encoded data in multiple clouds and perform computing tasks in a distributed way. In this work, nested MDS codes and the criterion of perfect secrecy are adopted. The allocation of encoded data to be stored in heterogeneous clouds with different computing capability is formulated as an optimization problem, subject to data reliability and security constraints. The problem is shown to be feasible if and only if the storage budget is above a certain level. The tradeoff between minimum computation time and storage budget is analytically characterized, and the former is proved to be a piecewise-linear decreasing function of the latter. When the storage budget increases beyond a certain threshold, the optimal computation time levels off. Closed-form expressions of minimum computation time and optimal storage allocation are obtained. Numerical results show that the optimized allocation outperforms equal allocation significantly if the computing rates of different clouds have large variation. Refereed/Peer-reviewed

Methodological handling of queries is a crucial requirement in social networks connected to a graph NoSQL database that incorporates massive amounts of data. The massive data need to be partitioned across numerous nodes so that the queries when executed can be retrieved from a parallel structure. A novel storage mechanism for effective query processing must to be established in graph databases for minimizing time overhead. This paper proposes a metaheuristic algorithm for partitioning of graph database across nodes by placement of all related information on same or adjacent nodes. The graph database allocation problem is proved to be NP-Hard. A metaheuristic algorithm comprising of Best Fit Decreasing with Ant Colony Optimization is proposed for data allocation in a distributed architecture of graph NoSQL databases. Lucene index is applied on proposed allocation for faster query processing. The proposed algorithm with Lucene is evaluated based on simulation results obtained from different heuristics available in literature.

Multiple constraints in SPMs are considered a problem that can be solved in a nondeterministic polynomial time. In this paper, we propose a novel approach solving the data allocations in multiple dimensional constraints. For supporting the approach, we develop a novel algorithm that is designed to solve the data allocations under multiple constraints in a polynomial time. Our proposed approach is a novel scheme of minimizing the total costs when executing SPM under multiple dimensional constraints. Our experimental evaluations have proved the adaptation of the proposed model that could be an efficient approach of solving data allocation problems for SPMs.

The need to design an optimally distributed database is increasingly important with the growth of information technology and computer networks. However, designing a distributed database is an extremely complex process due to a large number of geographically distributed sites and database relations. Moreover, decreasing communication costs and query response time should be taken into consideration. There are three main techniques applied to design a distributed database, namely Fragmentation, Data allocation, and Replication. It is notable that these techniques are often treated separately and rarely processed together. Some available allocation methods are applied regardless of how the fragmentation technique is performed or replication process is adopted. In contrast, other fragmentation techniques do not consider the allocation or the replication techniques. Therefore, the first and foremost step for designing an optimal database is to develop a comprehensive understanding of the current fragmentation, replication, and allocation techniques and their disadvantages. This article presents an attempt to fulfill this step by proposing a comprehensive taxonomy of the available fragmentation and allocation techniques in distributed database design. The article also discusses some case studies of these techniques for a deeper understanding of its achievements and limitations.

Scratchpad memory (SPM) is widely utilized in many embedded systems as a software- controlled on-chip memory to replace the traditional cache. New non-volatile memory (NVM) has emerged as a promising candidate to replace SRAM in SPM, due to its significant benefits, such as low-power consumption and high performance. In particular, several representative NVMs, such as PCM, ReRAM, and STT-RAM can build multiple-level cells (MLC) to achieve even higher density. Nevertheless, this triggers off higher energy overhead and longer access latency compared with its single-level cell (SLC) counterpart. To address this issue, this paper first proposes a specific SPM with morphable NVM, in which the memory cell can be dynamically programmed to the MLC mode or SLC mode. Considering the benefits of high-density MLC and low-energy SLC, a simple and novel optimization technique, named theory of thermal expansion and contraction, is presented to minimize the energy consumption and access latency in embedded systems. The basic idea is to dynamically adjust the size configure of SLC/MLC in SPM according to the different workloads of program and allocate the optimal storage medium for each data. Therefore, an integer linear programming formulation is first built to produce an optimal SLC/MLC SPM partition and data allocation. In addition, a corresponding approximation algorithm is proposed to achieve near-optimal results in polynomial time. Finally, the experimental results show that the proposed technique can effectively improve the system performance and reduce the energy consumption.

The effectiveness distributed database systems highly depends on the state of site that its task is to allocate fragments. This allocation purpose is performed for obtaining the minimum execute time and transaction cost of queries. There are some NP-hard problems that Data Allocation Problem (DAP) is one of them and solving this problem by means of enumeration method can be computationally expensive. Recently heuristic algorithms have been used to achieve desirable solutions. Due to fewer control parameters, robustness, speed convergence characteristics and easy adaptation to the problem, this paper propose a novel method based on Particle Swarm Optimization (PSO) algorithm which is suitable to minimize the total transmission cost for both the each site – fragment dependency and the each inter – fragment dependency. The core of the study is to solve DAP by utilizing and adaptation PSO algorithm, PSO-DAP for short. Allocation of fragments to the site has been done with PSO algorithm and its performance has been evaluated on 20 different test problems and compared with the state-of-art algorithms. Experimental results and comparisons demonstrate that proposed method generates better quality solutions in terms of execution time and total cost than compared state-of-art algorithms.

Non-Volatile Memory (NVM) has many advantages that provide new choice for embedded storage systems. However, the NVM has a limited number of writes and the overhead of write operation is greater than read operation. In order to overcome these drawbacks, it is necessary to reduce the read and write overhead of the NVM and make the NVM wear-leveling. In this paper, we use Iterational Optimal Data Allocation (IODA) algorithm to achieve the global minimum read and write overhead data allocation, and then, we propose Greedy and Global Balance Optimization Allocation (GBOA) algorithm to allocate NVM data into blocks to achieve NVM wear-leveling at the software-level. Experimental results show that the lifetime of NVM based on the Greedy algorithm is improved by 37.96% and the GBOA is improved by 49.52% compared to the Random algorithm on average, while we can obtain the global minimum read and write overhead.

Since the energy source is unstable in energy harvesting powered systems, checkpointing is a must for the energy harvesting powered systems. Non-volatile memory is used for keeping the persistence for the systems. However, it may also bring new problems for the systems, which are the inconsistency errors induced during program execution. In this work, we propose a checkpointing-aware data allocation method to reduce the total cost of checkpointing and program execution without the inconsistency errors. The experimental results show that the proposed method achieves 71.2% dynamic energy consumption reduction of checkpointing and program execution, and 9.9% reduction of total checkpointing and program execution time on average compared to the previous work without the inconsistency errors.

As the blockchain technology grows, the scale of the network is becoming huger and huger, which brings us a troublesome question: how to allocate our data into each computer of a blockchain network? For example, we can store our data from one computer, until the memory space filled, and then switch to another, spending a lot of time searching in sole computer. Or we can only put a small amount of data into one computer, resulting in a huge network and plenty of transmission time. In this paper, we compute the cost of all the allocations according to the mathematical model. By studying the formula, we get the solution of balancing the search time and the space occupancy by users' input. Based on the result, we design a data allocation strategy that can decide how many data allocated to an individual computer.

Solid State Disk (SSD), a type of NAND-flash memory semiconductor device becomes popular storage systems in various types of appliances. The surveys have brought out its essential demand for replacing the classical storage devices (e.g, Magnetic disk) in a multimedia streaming server such as a Video-On-Demand (VOD) server. The most important factor in providing an efficient server is the realtime playback experience by the users. However, the NAND-flash has three issues those related to the nature of its I/O accessing procedures. This paper studies the effectiveness of proper data layout in the SSD to overcome those issues in handling the VOD server real-time playback requirement. Data allocation acts as an engine that store or retrieve the continuous media (CM) data from the server and send to users at a proper rate, good quality and offer smooth VCR functions - pause, fast/slow rewind/forward. Two allocations those utilizing the rich parallelism stripping nature in SSD, namely; i) Sequential, and ii) Interleaving have been examined in this study. Both allocations intelligently organize the CM data using the residue class algorithm and the results from the simulation experiments shown both allocations have significant outcomes in terms of the access latency and the throughput.

High Performance Computing, which brings together all the players responsible for improving the computing performance of scientific applications on supercomputers, aims to achieve exaflopic performance. This race for performance is today characterized by the manufacture of heterogeneous machines in which each component is specialized. Among these components, system memories specialize too, and the trend is towards an architecture composed of several memories with complementary characteristics. The question arises then of these new machines use whose practical performance depends on the application data placement on the different memories. Compromising code update against performance is challenging. In this thesis, we have developed a data allocation on Heterogeneous Memory Architecture problem formulation. In this formulation, we have shown the benefit of a temporal analysis of the problem, because many studies were based solely on a spatial approach this result highlight their weakness. From this formulation, we developed an offline profiling tool to approximate the coefficients of the objective function in order to solve the allocation problem and optimize the allocation of data on a composite architecture composed of two main memories with complementary characteristics. In order to reduce the amount of code changes needed to execute an application according to our toolbox recommended allocation strategy, we have developed a tool that can automatically redirect data allocations from a minimum source code instrumentation. The performance gains obtained on mini-applications representative of the scientific applications coded by the community make it possible to assert that intelligent data allocation is necessary to fully benefit from heterogeneous memory resources. On some problem sizes, the gain between a naive data placement strategy, and an educated data allocation one, can reach up to ×3.75 speedup., Le Calcul Haute Performance, regroupant l’ensemble des acteurs responsables de l’amélioration des performances de calcul des applications scientifiques sur supercalculateurs, s’est donné pour objectif d’atteindre des performances exaflopiques. Cette course à la performance se caractérise aujourd’hui par la fabrication de machines hétérogènes dans lesquelles chaque composant est spécialisé. Parmi ces composants, les mémoires du système se spécialisent, et la tendance va vers une architecture composée de plusieurs mémoires aux caractéristiques complémentaires. La question se pose alors de l’utilisation de ces nouvelles machines dont la performance pratique dépend du placement des données de l’application sur les différentes mémoires. Dans cette thèse, nous avons développé une formulation du problème d’allocation de donnée sur une Architecture à Mémoires Hétérogènes. Dans cette formulation, nous avons fait apparaître le bénéfice que pourrait apporter une analyse temporelle du problème, parce que de nombreux travaux reposaient uniquement sur une approche spatiale. À partir de cette formulation, nous avons développé un outil de profilage hors ligne pour approximer les coefficients de la fonction objective afin de résoudre le problème d’allocation et d’optimiser l’allocation des données sur une architecture composée deux de mémoires principales aux caractéristiques complémentaires. Afin de réduire la quantité de modifications nécessaires pour prendre en compte la stratégie d’allocation recommandée par notre boîte à outils, nous avons développé un outil capable de rediriger automatiquement les allocations de données à partir d’un minimum d’instrumentation dans le code source. Les gains de performances obtenus sur des mini-applications représentatives des applications scientifiques codées par la communauté permet d’affirmer qu’une allocation intelligente des données est nécessaire pour bénéficier pleinement de ressources mémoires hétérogènes. Sur certaines tailles de problèmes, le gain entre un placement naïf est une allocation instruite peut atteindre un facteur ×3.75.

High Performance Computing, which brings together all the players responsible for improving the computing performance of scientific applications on supercomputers, aims to achieve exaflopic performance. This race for performance is today characterized by the manufacture of heterogeneous machines in which each component is specialized. Among these components, system memories specialize too, and the trend is towards an architecture composed of several memories with complementary characteristics. The question arises then of these new machines use whose practical performance depends on the application data placement on the different memories. Compromising code update against performance is challenging. In this thesis, we have developed a data allocation on Heterogeneous Memory Architecture problem formulation. In this formulation, we have shown the benefit of a temporal analysis of the problem, because many studies were based solely on a spatial approach this result highlight their weakness. From this formulation, we developed an offline profiling tool to approximate the coefficients of the objective function in order to solve the allocation problem and optimize the allocation of data on a composite architecture composed of two main memories with complementary characteristics. In order to reduce the amount of code changes needed to execute an application according to our toolbox recommended allocation strategy, we have developed a tool that can automatically redirect data allocations from a minimum source code instrumentation. The performance gains obtained on mini-applications representative of the scientific applications coded by the community make it possible to assert that intelligent data allocation is necessary to fully benefit from heterogeneous memory resources. On some problem sizes, the gain between a naive data placement strategy, and an educated data allocation one, can reach up to ×3.75 speedup.; Le Calcul Haute Performance, regroupant l’ensemble des acteurs responsables de l’amélioration des performances de calcul des applications scientifiques sur supercalculateurs, s’est donné pour objectif d’atteindre des performances exaflopiques. Cette course à la performance se caractérise aujourd’hui par la fabrication de machines hétérogènes dans lesquelles chaque composant est spécialisé. Parmi ces composants, les mémoires du système se spécialisent, et la tendance va vers une architecture composée de plusieurs mémoires aux caractéristiques complémentaires. La question se pose alors de l’utilisation de ces nouvelles machines dont la performance pratique dépend du placement des données de l’application sur les différentes mémoires. Dans cette thèse, nous avons développé une formulation du problème d’allocation de donnée sur une Architecture à Mémoires Hétérogènes. Dans cette formulation, nous avons fait apparaître le bénéfice que pourrait apporter une analyse temporelle du problème, parce que de nombreux travaux reposaient uniquement sur une approche spatiale. À partir de cette formulation, nous avons développé un outil de profilage hors ligne pour approximer les coefficients de la fonction objective afin de résoudre le problème d’allocation et d’optimiser l’allocation des données sur une architecture composée deux de mémoires principales aux caractéristiques complémentaires. Afin de réduire la quantité de modifications nécessaires pour prendre en compte la stratégie d’allocation recommandée par notre boîte à outils, nous avons développé un outil capable de rediriger automatiquement les allocations de données à partir d’un minimum d’instrumentation dans le code source. Les gains de performances obtenus sur des mini-applications représentatives des applications scientifiques codées par la communauté permet d’affirmer qu’une allocation intelligente des données est nécessaire pour bénéficier pleinement de ressources mémoires hétérogènes. Sur certaines tailles de problèmes, le gain entre un placement naïf est une allocation instruite peut atteindre un facteur ×3.75.

Synchronize processing is a major challenge in multicore processor units. In the designing of multicore processor, as the processing unit work simultaneously, the synchronization of data, instruction and the delay encountered is to be minimized. In the effort for providing synchronous operation, delay synchronization is the main criterion. The delay encountered in bus switching, data fetching, instruction/data allocation, and resource allocation delay is a major concern. In the minimization of delay constraint in resource allocation, the mesochronous operation is suggested. In this work, an interface to mesochronous operation for multicore processor operation is proposed. The presented approach defines the delay application in multiprocessor application with delay synchronization.

Multi-socket machines with 1-100 TBs of physical memory are becoming prevalent. Applications running on multi-socket machines suffer non-uniform bandwidth and latency when accessing physical memory. Decades of research have focused on data allocation and placement policies in NUMA settings, but there have been no studies on the question of how to place page-tables amongst sockets. We make the case for explicit page-table allocation policies and show that page-table placement is becoming crucial to overall performance. We propose Mitosis to mitigate NUMA effects on page-table walks by transparently replicating and migrating page-tables across sockets without application changes. This reduces the frequency of accesses to remote NUMA nodes when performing page-table walks. Mitosis uses two components: (i) a mechanism to enable efficient page-table replication and migration; and (ii) policies for processes to efficiently manage and control page-table replication and migration. We implement Mitosis in Linux and evaluate its benefits on real hardware. Mitosis improves performance for large-scale multi-socket workloads by up to 1.34x by replicating page-tables across sockets. Moreover, it improves performance by up to 3.24x in cases when the OS migrates a process across sockets by enabling cross-socket page-table migration.

Phase change memory (PCM) is considered as a promising alternative of DRAM-based main memory in embedded systems. A PCM cell can be dynamically programmed to be in either multiple-level cell(MLC) mode or single-level cell(SLC) mode. With this morphable feature, we can utilize the high-density of MLC and low-latency of SLC, to satisfy various memory requirements of specific applications in embedded systems. However, compared to its SLC counterpart, the lifetime of MLC is limited.To address this issue, this paper proposes a simple and effective wear-leveling technique, named Mixer, to enhance the lifetime of morphable PCM considering the program specific features. We first build an Integer Linear Programming(ILP) formulation to dynamically configure the optimal SLC/MLC partition in morphable PCM, and produce the best data allocation for each variable to achieve a balanced write distribution in morphable PCM with low memory access cost. The basic idea is to allocate low-latency SLC and high-density MLC cells for write intensive variables and other ordinary variables, respectively. We then propose a polynomial time algorithm to achieve near-optimal results. The evaluation results show that the proposed technique can effectively improve the lifetime of morphable PCM in embedded systems compared with previous work.

Thanks to the advance of cloud computing technologies, users can access the data stored at cloud data centers at any time and from any where. However, the data centers are usually sparsely distributed over the Internet and are far away from end users. In this paper, we consider to construct a cache network by a large number of cache nodes close to the end users in order to minimize the data access delay.We firstly formulate the problem of placing the replicas of data items to cache nodes as a mixed integer programming (MIP) problem. Then, we proposed an efficient heuristic algorithm that allocates at least one replica of each data item in the cache network and attempt to allocate more data items so as to minimize the total data access cost. The simulation results show that our proposed algorithm behaves much better than a well-known LRU algorithm and the computation complexity is limited.

The paper presents a scalable distributed algorithm for static and dynamic data allocation in LuNA fragmented programming system. LuNA is intended for automation of construction of parallel programs, which implement large-scale numerical models for multicomputers with large number of computing nodes. The proposed algorithm takes into account data structure of the numerical model implemented, provides static and dynamic load balancing and can be used with various network topologies.

In cloud computing, server consolidation is the part where very few persons go through the same. By consolidating the unused server space, memory can be reused for another data allocation. The objective of this paper is to improve resource utilization. Residual resource fragmentation refers to the state of the data center where sufficient amount of residual resources are available for any new VM allocation. To achieve this, there are three methods followed here. Active physical servers are identified. Then the maximum utilization of the resources is found out. Finally the resources are allocated and scheduled using the developed algorithm. In this work, we have proposed a new algorithm enhanced cloud consolidating algorithm. This algorithm improves some of the qualities of the cloud consolidating algorithm. Here the allocation technique is based on the cost and the memory.

In this work, as part of our continuous effort, an optimized heuristic technique is proposed. The basic aim of this technique to fragment data vertically in Distributed Database System (DDBS). Drive by the queries along with using hierarchical clustering algorithm, the proposed technique seeks to propose an aggregated similarity measure to properly perform the clustering-based vertical data fragmentation. Data replication and allocation are also investigated to produce a comprehensive solution. As a matter of fact, the key concern is to find an effective best-fitting solution for improving DDBS throughput through developing an aggregated similarity based data fragmentation process, drawing a site clustering algorithm, and presenting a greedy-based transmission costs-reducing data allocation algorithm. Moreover, data replication is carefully considered in such a way that cost is to be essentially minimized.

This paper details out the application of a predictive analysis tool to ‘S’ Field's commingled production, aiming to enhance production allocation and reservoir understanding without the need of well intervention and a reduced frequency of zonal rate tests and data acquisition. Allocation of the production data to its respective reservoirs is performed via a novel Multi-Phase Allocation method (MPA), taking into account the water production trending evolution derived from relative permeability behavior of oil-water in each reservoir to compute flow rates for liquid phases over time. The precision of the derived rates is constrained by actual zonal rates tests through Inflow Control Valves (ICVs). This method will be cross referenced against ‘S’ Field's existing zonal rate calculation algorithm, utilizing input data from well tests results and real time pressure and temperature data. The MPA method demonstrates improvement in the allocation of production data as compared to the conventional KH-methodology as MPA takes into account the water cut trending between reservoirs. Leveraging on ICVs to obtain actual zonal rate measurements, this greatly reduces the range of uncertainty in the allocation process. MPA derived production split ratios closely match the split ratios derived from the ‘S’ Field's existing zonal rate calculation algorithm, which utilizes input data from well tests results and real time pressure and temperature data from down hole gauges. It is observed that the usage of actual measured zonal rate tests reduces the range of uncertainty of the MPA data. A combination of novel multi-phase deliverability models coupled with smart field technologies such as intelligent completions and real-time surveillance and analysis tools will increase the accuracy of the back allocation of multi-phase production data in commingled reservoirs.

Charge trap (CT) based 3D NAND flash is predominating the flash storage market due to higher density, better performance and endurance than planar flash. CT-based 3D flash programs multiple pages in a word line at a time, called one-shot programming, unlike planar flash which programs one page at a time. Solid state drives (SSDs) utilize the internal parallelism to improve the performance, but one-shot programming is likely to program logically sequential data into one parallel unit (i.e., a plane) and thus degrades the read parallelism. In this paper, we propose a one-shot programming aware data allocation policy, called OSPADA, to improve the read performance of CT flash based SSDs by enhancing read parallelism. OSPADA reorders written data to distribute logically sequential data into different parallel units using the distance aware round-robin strategy. Experimental results show that OSPADA improves the read performance by up to 22.8% compared with traditional dynamic data allocation policies.

The limited computational and storage capabilities of the devices interconnected in Internet of Things (IoT) make them to host only a sub-set of the the collected data. Every device, i.e., an IoT node, should keep only the necessary data locally, thus, it can be able to process them and provide responses in limited time. Nodes can act as a team and cooperate to store the data close to the processing of tasks defined in the form of queries. In this paper, we propose a model for deciding the allocation of data in a set of IoT nodes. Every node decides if the observed data are correlated with the available datasets or they are outliers. We propose an ensemble scheme for multidimensional outliers detection that results, in real time, the final decision. When data are accepted to be locally stored, nodes select their peers where data will be replicated. This way, we keep the data in multiple locations in the network aiming to reduce latency in the provision of responses and support a fault tolerant mechanism. The replication decision is based on the correlation of the incoming data with the present datasets. We analytically describe our model and evaluate it through extensive simulations presenting its pros and cons.

More and more Non-Volatile Memories (NVM) have been widely applied to various embedded systems to build the heterogeneous memory architecture. However, the write-endurance of NVM remains a great challenge. Hence, we should take full consideration of the write-endurance of NVM when allocating data on heterogeneous memory architecture. There is an observation that, for most real workloads, about 10% of data account for 90% write operations. This brings us an opportunity to reduce the write wear of NVM through carefully allocating write-intensive data. In this paper, we explore the problem that how to find a balance between the system cost and write-endurance of NVM for data allocation on heterogeneous memory architecture. We propose a write-aware data allocation algorithm, WADA. WADA can not only greatly reduce the write wear of NVM, but also guarantee the near-optimal system cost. We also propose an integer linear programming (ILP) model to generate an optimal data allocation, which can obtain the minimum cost. The result of ILP can be used as a standard to evaluate the efficiency of other algorithms. Experiments show that WADA outperforms all the other algorithms on both system cost and write wear of NVM. Compared to previous algorithms, WADA can reduce up to 47.77% system cost and 60.89% write wear of NVM. Compared to ILP, WADA can achieve the near-optimal system cost within just 2% difference.