A5088654262- Advanced Data Storage Technologies
- Cloud Computing and Resource Management
- Parallel Computing and Optimization Techniques
- Distributed systems and fault tolerance
- Software-Defined Networks and 5G
- Advanced Database Systems and Queries
- Distributed and Parallel Computing Systems
- Data Management and Algorithms
- Algorithms and Data Compression
- Modular Robots and Swarm Intelligence
- Caching and Content Delivery
- Semantic Web and Ontologies
- Interconnection Networks and Systems
- Semiconductor materials and devices
- Software System Performance and Reliability
- Manufacturing Process and Optimization
- Service-Oriented Architecture and Web Services
- Complex Systems and Time Series Analysis
- Stock Market Forecasting Methods
- Graph Theory and Algorithms
- Music and Audio Processing
- Blind Source Separation Techniques
- Time Series Analysis and Forecasting
- Mobile Agent-Based Network Management
- Engineering Technology and Methodologies
University of Fribourg
2020-2024
New York University
2004
Design Business Association
2003
Télécom Paris
2003
Recording sensor data is seldom a perfect process. Failures in power, communication or storage can leave occasional blocks of missing, affecting not only real-time monitoring but also compromising the quality near- and off-line analysis. Several recovery (imputation) algorithms have been proposed to replace missing blocks. Unfortunately, little known about their relative performance, as existing comparisons are limited either small subset relevant very few datasets often both. Drawing...
The trend toward specialized processing devices such as TPUs, DPUs, GPUs, and FPGAs has exposed the weaknesses of PCIe in interconnecting these their hosts. Several attempts have been proposed to improve, augment, or downright replace PCIe, more recently, efforts converged into a standard called Compute Express Link (CXL). CXL is already on version 2.0 terms commercial availability, but its potential radically change conventional server architecture only just started surface. For example,...
Programmable switches based on the Protocol Independent Switch Architecture (PISA) have greatly enhanced flexibility of today's networks by allowing new packet protocols to be deployed without any hardware changes. They also been instrumental in enabling a computing paradigm which parts an application's logic run within network core (in-network computing).
Transaction logging and log shipping are standard techniques to provide recoverability high availability in data management systems. They entail an update a local file at every transaction sending such remote site coordinated fashion. Modern databases have leveraged technologies as Persistent Memory (PM) RDMA-enabled networking perform these updates fast possible. This mix of technologies, however, presents several drawbacks: some not portable, restricting deployments single class machines;...
This article describes some of the ongoing research projects related to structured data management at Google today. The organization encourages scientists work closely with engineering teams. As a result, tend be motivated by real needs faced Google's products and services, solutions are put into production tested rapidly. In addition, because sheer scale which operates, challenges services often require innovations.
Graph Pattern Mining (GPM) is a class of algorithms that identifies given shapes within graph, e.g., cliques certain size. Any area graph can contain shape interest, but in real-world graphs, these tend to be concentrated areas deemed skewed. Because mining skewed dominate GPM computations, the overwhelming majority state-of-the-art techniques break such into many small parts and load balance them across servers. This paper takes diametrically opposite approach: we suggest framework...
We introduce Transaction Triaging, a set of techniques that manipulate streams transaction requests and responses while they travel to from database server. Compared normal streams, the triaged ones execute faster once reach database. The triaging algorithms do not interfere with execution nor require adherence any particular concurrency control method, making them easy port across systems. Triaging leverages recent programmable networking hardware can perform computations on in-flight data....
Gone is the time when a Solid-State Drive (SSD) was just fast drop-in replacement for Hard-Disk (HDD). Thanks to NVMe ecosystem, nowadays, SSDs are accessed through specific interfaces and modern I/O frameworks. have also grown versatile with can now support various use cases ranging from cold, high-density storage hot, low-latency ones. The body of knowledge about building such different devices mostly available, but it less than accessible non-experts. Finding which device variation better...
Financial time series streams are watched closely by millions of traders. What exactly do they look for and how can we help them it faster? Physicists study the emerging from their sensors. The same question holds them. Musicians produce series. Consumers may want to compare This tutorial presents techniques case studies four problems:1. Finding sliding window correlations in financial, physical, other applications.2. Discovering bursts large sensor data gamma rays.3. Matching hums recorded...
Modern cloud networks are a fundamental pillar of data-intensive applications. They provide high-speed transaction (packet) rates and low overhead, enabling, for instance, truly scalable database designs. These networks, however, fundamentally different from conventional ones. Arguably, the two key discerning technologies RDMA programmable network devices. Today, these not niche anymore widely deployed across all major vendors. The question is thus if but how new breed applications can...
Modern DBMS engines can achieve unprecedented transaction processing speeds thanks to the invention of clever data structures, concurrency schemes, and improvements in CPU memory subsystems. However, developing realistic efficient networked clients benchmark these systems remains daunting. Simply put, traditional client-side networking stacks present high overheads thus cannot exercise performance that modern DBMSs can, principle, provide. In this demo, we propose a different approach...
The growing trend towards specialization has led to a proliferation of accelerators and alternative processing devices. When embedded in conventional computer architectures, the PCIe link connecting CPU these devices becomes bottleneck. Several proposals for designs have been put forward, with efforts having now converged into Compute Express Link (CXL) specification. CXL is an interconnect protocol on top more modern powerful interface. While still version 1.0 terms commercial availability,...