Heterogeneous architecture for the optimization of large-scale graph processing data centres

Con el auge del Cloud Computing, las aplicaciones de proceso de datos han sufrido un incremento de demanda, y por ello ha cobrado importancia lograr m�ás eficiencia en los Centros de Proceso de datos. El objetivo de este trabajo es la obtenci�ón de herramientas que permitan analizar la viabilidad y rentabilidad de diseñar Centros de Datos especializados para procesamiento de datos, con una arquitectura, sistemas de refrigeraci�ón, etc. adaptados. Algunas aplicaciones de procesamiento de datos se benefician de las arquitecturas software, mientras que en otras puede ser m�ás eficiente un procesamiento con arquitectura hardware. Debido a que ya hay software con muy buenos resultados en el procesamiento de grafos, como el sistema XPregel, en este proyecto se realizará una arquitectura hardware en VHDL, implementando el algoritmo PageRank de Google de forma escalable. Se ha escogido este algoritmo ya que podr��á ser m�ás eficiente en arquitectura hardware, debido a sus características concretas que se indicaráan m�ás adelante. PageRank sirve para ordenar las p�áginas por su relevancia en la web, utilizando para ello la teorí��a de grafos, siendo cada página web un vértice de un grafo; y los enlaces entre páginas, las aristas del citado grafo. En este proyecto, primero se realizará un an�álisis del estado de la técnica. Se supone que la implementaci�ón en XPregel, un sistema de procesamiento de grafos, es una de las m�ás eficientes. Por ello se estudiará esta �ultima implementaci�ón. Sin embargo, debido a que Xpregel procesa, en general, algoritmos que trabajan con grafos; no tiene en cuenta ciertas caracterí��sticas del algoritmo PageRank, por lo que la implementaci�on no es �optima. Esto es debido a que en PageRank, almacenar todos los datos que manda un mismo v�értice es un gasto innecesario de memoria ya que todos los mensajes que manda un vértice son iguales entre sí e iguales a su PageRank. Se realizará el diseño en VHDL teniendo en cuenta esta caracter��ística del citado algoritmo,evitando almacenar varias veces los mensajes que son iguales. Se ha elegido implementar PageRank en VHDL porque actualmente las arquitecturas de los sistemas operativos no escalan adecuadamente. Se busca evaluar si con otra arquitectura se obtienen mejores resultados. Se realizará un diseño partiendo de cero, utilizando la memoria ROM de IPcore de Xillinx (Software de desarrollo en VHDL), generada autom�áticamente. Se considera hacer cuatro tipos de módulos para que as�� el procesamiento se pueda hacer en paralelo. Se simplificar�á la estructura de XPregel con el fin de intentar aprovechar la particularidad de PageRank mencionada, que hace que XPregel no le saque el m�aximo partido. Despu�és se escribirá el c�ódigo, realizando una estructura escalable, ya que en la computación intervienen millones de páginas web. A continuación, se sintetizar�á y se probará el código en una FPGA. El �ultimo paso será una evaluaci�ón de la implementaci�ón, y de posibles mejoras en cuanto al consumo.

A Scalable General Purpose System for Large-Scale Graph Processing

Graph analytics is an important and computationally demanding class of data analytics. It is essential to balance scalability, ease-of-use and high performance in large scale graph analytics. As such, it is necessary to hide the complexity of parallelism, data distribution and memory locality behind an abstract interface. The aim of this work is to build a scalable graph analytics framework that does not demand significant parallel programming experience based on NUMA-awareness.
The realization of such a system faces two key problems:
(i)~how to develop a scale-free parallel programming framework that scales efficiently across NUMA domains; (ii)~how to efficiently apply graph partitioning in order to create separate and largely independent work items that can be distributed among threads.

An improved memory management scheme for large scale graph computing engine GraphChi

GraphChi is the first reported disk-based graph engine that can handle billion-scale graphs on a single PC efficiently. GraphChi is able to execute several advanced data mining, graph mining and machine learning algorithms on very large graphs. With the novel technique of parallel sliding windows (PSW) to load subgraph from disk to memory for vertices and edges updating, it can achieve data processing performance close to and even better than those of mainstream distributed graph engines. GraphChi mentioned that its memory is not effectively utilized with large dataset, which leads to suboptimal computation performances. In this paper we are motivated by the concepts of 'pin ' from TurboGraph and 'ghost' from GraphLab to propose a new memory utilization mode for GraphChi, which is called Part-in-memory mode, to improve the GraphChi algorithm performance. The main idea is to pin a fixed part of data inside the memory during the whole computing process. Part-in-memory mode is successfully implemented with only about 40 additional lines of code to the original GraphChi engine. Extensive experiments are performed with large real datasets (including Twitter graph with 1.4 billion edges). The preliminary results show that Part-in-memory mode memory management approach effectively reduces the GraphChi running time by up to 60% in PageRank algorithm. Interestingly it is found that a larger portion of data pinned in memory does not always lead to better performance in the case that the whole dataset cannot be fitted in memory. There exists an optimal portion of data which should be kept in the memory to achieve the best computational performance.

A big data platform for large scale event processing

To date, big data applications have focused on the store-and-process paradigm. In this paper we describe an initiative to deal with big data applications for continuous streams of events. In many emerging applications, the volume of data being streamed is so large that the traditional ‘store-then-process’ paradigm is either not suitable or too inefficient. Moreover, soft-real time requirements might severely limit the engineering solutions. Many scenarios fit this description. In network security for cloud data centres, for instance, very high volumes of IP packets and events from sensors at firewalls, network switches and routers and servers need to be analyzed and should detect attacks in minimal time, in order to limit the effect of the malicious activity over the IT infrastructure. Similarly, in the fraud department of a credit card company, payment requests should be processed online and need to be processed as quickly as possible in order to provide meaningful results in real-time. An ideal system would detect fraud during the authorization process that lasts hundreds of milliseconds and deny the payment authorization, minimizing the damage to the user and the credit card company.

Large scale data processing using MapReduce

As massive data sets become increasingly available, people are facing the problem of how to effectively process and understand these data. Traditional sequential computing models are giving way to parallel and distributed computing models, such as MapReduce, both due to the large size of the data sets and their high dimensionality. This dissertation, as in the same direction of other researches that are based on MapReduce, tries to develop effective techniques and applications using MapReduce that can help people solve large-scale problems. Three different problems are tackled in the dissertation. The first one deals with processing terabytes of raster data in a spatial data management system. Aerial imagery files are broken into tiles to enable data parallel computation. The second and third problems deal with dimension reduction techniques that can be used to handle data sets of high dimensionality. Three variants of the nonnegative matrix factorization technique are scaled up to factorize matrices of dimensions in the order of millions in MapReduce based on different matrix multiplication implementations. Two algorithms, which compute CANDECOMP/PARAFAC and Tucker tensor decompositions respectively, are parallelized in MapReduce based on carefully partitioning the data and arranging the computation to maximize data locality and parallelism.

Dithen: A Computation-as-a-Service Cloud Platform For Large-Scale Multimedia Processing

We present Dithen, a novel computation-as-a-service (CaaS) cloud platform specifically tailored to the parallel ex-ecution of large-scale multimedia tasks. Dithen handles the upload/download of both multimedia data and executable items, the assignment of compute units to multimedia workloads, and the reactive control of the available compute units to minimize the cloud infrastructure cost under deadline-abiding execution. Dithen combines three key properties: (i) the reactive assignment of individual multimedia tasks to available computing units according to availability and predetermined time-to-completion constraints; (ii) optimal resource estimation based on Kalman-filter estimates; (iii) the use of additive increase multiplicative decrease (AIMD) algorithms (famous for being the resource management in the transport control protocol) for the control of the number of units servicing workloads. The deployment of Dithen over Amazon EC2 spot instances is shown to be capable of processing more than 80,000 video transcoding, face detection and image processing tasks (equivalent to the processing of more than 116 GB of compressed data) for less than $1 in billing cost from EC2. Moreover, the proposed AIMD-based control mechanism, in conjunction with the Kalman estimates, is shown to provide for more than 27% reduction in EC2 spot instance cost against methods based on reactive resource estimation. Finally, Dithen is shown to offer a 38% to 500% reduction of the billing cost against the current state-of-the-art in CaaS platforms on Amazon EC2 (Amazon Lambda and Amazon Autoscale). A baseline version of Dithen is currently available at dithen.com.

Identifying rate-limiting nodes in large-scale cortical networks for visuospatial processing: An illustration using fMRI

With the advent of functional neuroimaging techniques, in particular functional magnetic resonance imaging (fMRI), we have gained greater insight into the neural correlates of visuospatial function. However, it may not always be easy to identify the cerebral regions most specifically associated with performance on a given task. One approach is to examine the quantitative relationships between regional activation and behavioral performance measures. In the present study, we investigated the functional neuroanatomy of two different visuospatial processing tasks, judgement of line orientation and mental rotation. Twenty-four normal participants were scanned with fMRI using blocked periodic designs for experimental task presentation. Accuracy and reaction time (RT) to each trial of both activation and baseline conditions in each experiment was recorded. Both experiments activated dorsal and ventral visual cortical areas as well as dorsolateral prefrontal cortex. More regionally specific associations with task performance were identified by estimating the association between (sinusoidal) power of functional response and mean RT to the activation condition; a permutation test based on spatial statistics was used for inference. There was significant behavioral-physiological association in right ventral extrastriate cortex for the line orientation task and in bilateral (predominantly right) superior parietal lobule for the mental rotation task. Comparable associations were not found between power of response and RT to the baseline conditions of the tasks. These data suggest that one region in a neurocognitive network may be most strongly associated with behavioral performance and this may be regarded as the computationally least efficient or rate-limiting node of the network.

New graph-based algorithms to efficiently solve large scale open pit mining optimisation problems

In the mining optimisation literature, most researchers focused on two strategic-level and tactical-level open-pit mine optimisation problems, which are respectively termed ultimate pit limit (UPIT) or constrained pit limit (CPIT). However, many researchers indicate that the substantial numbers of variables and constraints in real-world instances (e.g., with 50-1000 thousand blocks) make the CPIT’s mixed integer programming (MIP) model intractable for use. Thus, it becomes a considerable challenge to solve the large scale CPIT instances without relying on exact MIP optimiser as well as the complicated MIP relaxation/decomposition methods. To take this challenge, two new graph-based algorithms based on network flow graph and conjunctive graph theory are developed by taking advantage of problem properties. The performance of our proposed algorithms is validated by testing recent large scale benchmark UPIT and CPIT instances’ datasets of MineLib in 2013. In comparison to best known results from MineLib, it is shown that the proposed algorithms outperform other CPIT solution approaches existing in the literature. The proposed graph-based algorithms leads to a more competent mine scheduling optimisation expert system because the third-party MIP optimiser is no longer indispensable and random neighbourhood search is not necessary.

Divide and Conquer: Efficient Large-Scale Structure from Motion Using Graph Partitioning

Despite significant advances in recent years, structure-from-motion (SfM) pipelines suffer from two important drawbacks. Apart from requiring significant computational power to solve the large-scale computations involved, such pipelines sometimes fail to correctly reconstruct when the accumulated error in incremental reconstruction is large or when the number of 3D to 2D correspondences are insufficient. In this paper we present a novel approach to mitigate the above-mentioned drawbacks. Using an image match graph based on matching features we partition the image data set into smaller sets or components which are reconstructed independently. Following such reconstructions we utilise the available epipolar relationships that connect images across components to correctly align the individual reconstructions in a global frame of reference. This results in both a significant speed up of at least one order of magnitude and also mitigates the problems of reconstruction failures with a marginal loss in accuracy. The effectiveness of our approach is demonstrated on some large-scale real world data sets.

Large scale heterogeneous networks, the Davis-Wielandt shell, and graph separation

We consider a large scale network of interconnected heterogeneous dynamical components. Scalable stability conditions are derived that involve the input/output properties of individual subsystems and the interconnection matrix. The analysis is based on the Davis-Wielandt shell, a higher dimensional version of the numerical range with important convexity properties. This can be used to allow heterogeneity in the agent dynamics while relaxing normality and symmetry assumptions on the interconnection matrix. The results include small gain and passivity approaches as special cases, with the three dimensional shell shown to be inherently connected with corresponding graph separation arguments. © 2012 Society for Industrial and Applied Mathematics.

Optimum-Path Forest Classifier for Large Scale Biometric Applications

This paper addresses biometric identification using large databases, in particular, iris databases. In such applications, it is critical to have low response time, while maintaining an acceptable recognition rate. Thus, the trade-off between speed and accuracy must be evaluated for processing and recognition parts of an identification system. In this paper, a graph-based framework for pattern recognition, called Optimum-Path Forest (OPF), is utilized as a classifier in a pre-developed iris recognition system. The aim of this paper is to verify the effectiveness of OPF in the field of iris recognition, and its performance for various scale iris databases. The existing Gauss-Laguerre Wavelet based coding scheme is used for iris encoding. The performance of the OPF and two other - Hamming and Bayesian - classifiers, is compared using small, medium, and large-scale databases. Such a comparison shows that the OPF has faster response for large-scale databases, thus performing better than the more accurate, but slower, classifiers.

Quality of Service in Distributed Stream Processing for large scale Smart Pervasive Environments

The wide diffusion of cheap, small, and portable sensors integrated in an unprecedented large variety of devices and the availability of almost ubiquitous Internet connectivity make it possible to collect an unprecedented amount of real time information about the environment we live in. These data streams, if properly and timely analyzed, can be exploited to build new intelligent and pervasive services that have the potential of improving people's quality of life in a variety of cross concerning domains such as entertainment, health-care, or energy management. The large heterogeneity of application domains, however, calls for a middleware-level infrastructure that can effectively support their different quality requirements. In this thesis we study the challenges related to the provisioning of differentiated quality-of-service (QoS) during the processing of data streams produced in pervasive environments. We analyze the trade-offs between guaranteed quality, cost, and scalability in streams distribution and processing by surveying existing state-of-the-art solutions and identifying and exploring their weaknesses. We propose an original model for QoS-centric distributed stream processing in data centers and we present Quasit, its prototype implementation offering a scalable and extensible platform that can be used by researchers to implement and validate novel QoS-enforcement mechanisms. To support our study, we also explore an original class of weaker quality guarantees that can reduce costs when application semantics do not require strict quality enforcement. We validate the effectiveness of this idea in a practical use-case scenario that investigates partial fault-tolerance policies in stream processing by performing a large experimental study on the prototype of our novel LAAR dynamic replication technique. Our modeling, prototyping, and experimental work demonstrates that, by providing data distribution and processing middleware with application-level knowledge of the different quality requirements associated to different pervasive data flows, it is possible to improve system scalability while reducing costs.

Listening to nature : techniques for large-scale monitoring of ecosystems using acoustics

Climate change and human activity are subjecting the environment to unprecedented rates of change. Monitoring these changes is an immense task that demands new levels of automated monitoring and analysis. We propose the use of acoustics as a proxy for the time consuming auditing of fauna, especially for determining the presence/absence of species. Acoustic monitoring is deceptively simple; seemingly all that is required is a sound recorder. However there are many major challenges if acoustics are to be used for large scale monitoring of ecosystems. Key issues are scalability and automation. This paper discusses our approach to this important research problem. Our work is being undertaken in collaboration with ecologists interested both in identifying particular species and in general ecosystem health.