Reda will present his recent work in Distributed Collaborative Filtering on Friday 10 Oct at 3:30pm (room to be defined). A Distributed Collaborative Filtering Algorithm With Multiple and Heterogeneous Data Sources. Recommender systems are used as a mean to supply users with content that may be of interest to them. They have attracted the attention of the research community, and have become a popular research topic, where many aspects and dimensions have been studied to make them more accurate and effective (this includes the: social dimension, geographical dimension, diversification aspect, etc.). Collaborative filtering (CF) is certainly one of the most famous recommendation methods, which consists in predicting whether, or how much, a user will like (or dislike) an item by leveraging knowledge of that user’s preferences as well as those of other users. However, in practice, users interact and express their opinion on only a small subset of items, which makes the corresponding user-item rating matrix very sparse. Consequently, in a recommender system, this data sparsity induced mainly two problems: (1) the lack of data to effectively model users’ preferences (news users suffer from the cold-start problem), and similarly (2) the lack of data to effectively model items’ preferences (new items suffer from the cold-start problem since no user has rated them). However, on the other hand, users use many online services, which can provide information about their interest and the content of items (e.g. Google search engine, Facebook, Twitter, etc). These services may be valuable data sources, which supply information to help a recommender system in modeling users and items’ preferences, and thus, make the recommender system more precise. Moreover, these data sources are distributed, and geographically distant from each other, which raise many research problems and challenges to design a distributed recommendation algorithm. Hence, in this talk, we present a new distributed collaborative filtering algorithm, which exploits and combine these multiple and heterogeneous data sources to improve the recommendation quality. Short bio: Reda Bouadjenek received a master and a PhD degree in computer science from the University of Versailles, France, in 2009 and 2013 respectively. He is currently a postdoctoral researcher at INRIA, and works on recommender systems. Previously, he worked for Alcatel-Lucent Bell Labs France from 2010 to 2013 as researcher, then was a visitor researcher at NICTA&ANU, Australia, in 2013. His research interests include Information Retrieval, Social Network Analysis, Data Mining, Machine Learning, Recommender Systems, and Databases.

Monday 30th june 2014, in MONTPELLIER, 95 rue de la Galéra

IBC & LIRMM (UM2, CNRS-Mastodons), INRIA-UCSB associated team Bigdatanet

Organisation : Esther Pacitti Lirmm reception desk : +33 (0)4 67 41 85 85

Workshop Objective

In the context of the Mastodons project in Montpellier, we are addressing problems related to the management and analysis of big scientific data, in particular biology data such as those produced by next generation sequencing tools and plant phenotyping platforms. The objective of the workshop is to discuss emerging solutions for big data management with world-class scientists. More information on Mastodons Web Site.

Salle Galera 127 le 23-06-2014, 11h
Organisé par l’équipe Zenith

In-Memory Analytics: Accelerating Business Performance
Antoine Chambille, Romain Colle
QuartetFS, Paris

The Big Data trend is a rebirth for Business Intelligence. On the one hand the web companies use technologies like Hadoop to extract value from data previously out of reach, because it is too big or because it is not structured. On the other side the new databases that store the data in-memory reach levels of performance such that they can perform complex and interactive analysis on live data that changes in real-time.

In concrete terms those In-Memory databases are the foundation for a new generation of business applications that bring the power of analytics to the hands of the decision makers who “run the business”.

Through this crash course on In-Memory technology, we will see through practical examples the competitive advantage it already brings to the best-performing organizations in the fields of e-commerce, logistics and finance.

About the speakers

Antoine Chambille is Head of Research and Development at Quartet FS. He joined Quartet FS soon after its creation back in 2005 and has been leading the team in charge of designing, developing and supporting Quartet FS’s in-memory analytics solutions. As one of the first employees, Antoine was heavily involved in the design of ActivePivot Server, Quartet FS’ in-memory OLAP engine.
Before joining Quartet FS, Antoine worked several years for a consulting firm specialised in the financial sector. From his years in consulting, he developed a strong customer orientation and he is keen on keeping a close eye on customers’ use cases. Antoine graduated from Ecole Polytechnique and Telecom Paris.

Romain Colle is a Project Manager within the Research and Development team of Quartet FS. He designed and developed Sentinel, QuartetFS’ flagship monitoring solution, and led the development efforts on ActivePivot’s Distributed Architecture. Romain has been involved in large ActivePivot projects such as Societe Generale in Paris, J.P.Morgan in London and DekaBank in Frankfurt. He joined Quartet FS in 2010 after 3 years spent at Oracle’s Headquarters in Redwood Shores where he contributed to the development of their core database. Romain is graduated from the Stanford University (he holds a Master of Science in Computer Science) and from “Centrale” in Paris”.

About Quartet FS and ActivePivot

Quartet FS provides business users with instant insight into massive amounts of data streaming at high frequency for timely and context-aware decision-making. Using Quartet FS’ in-memory aggregation engine ActivePivot, organisations are able to build 24×7, sense-and-respond applications that help them accelerate business performance, optimize operations, reduce operational risk and react to the unexpected.

Created in 2005, Quartet FS is a privately owned company with offices in Paris, London, New York, Hong Kong and Singapore. With more than fifty live implementations in large international groups, the company serves customers operating in time-sensitive and data-intensive environments such as financial services, market exchanges, logistics, transportation, and retail.

ActivePivot allows business users to be able to extract actionable intelligence from massive amounts of fast moving data – enabling them to make informed decisions on the spot. An in-memory Hybrid Transactional and Analytics Processing engine, ActivePivot aggregates data from multiple sources and processes multi-dimensional queries at unparalleled speeds on data that is updated on the fly.
As a result, business users are able to:

• Focus on what really matters by pinpointing anomalies at an early stage with ActivePivot detecting changes in business conditions and pushing meaningful alerts
• Get answers in sub-second time to any analytical enquiry with ActivePivot processing millions of live records across systems using its in-memory aggregation capabilities
• Drill-down and view data from any angle, at any point in the past, with ActivePivot providing the required contextual information for root cause analysis.
• Run ‘What-if’ analysis and evaluate the effect of alternative scenarios on the business

Zenith Seminar Room Galera 127 on May  28, 2014, 11am. CLyDE Mid-Flight: What we have learnt so far about the SSD-Based IO Stack  Philippe Bonnet, INRIA and IT University of Copenhagen Abstract: The quest for energy proportional systems and the growing performance gap between processors and magnetic disks has led to the adoption of SSDs as secondary storage of choice for a large range of systems.  Indeed, SSDs offer great performance (tens of flash chips wired in parallel can deliver hundreds of thousands accesses per second) with low energy consumption. This evolution introduces a mismatch between the simple disk model that underlies the design of today’s database systems and the complex SSDs of today’s computers. This mismatch leads to unpredictable performance, with orders of magnitude slowdown in IO latency that can hit an application anytime. To attack this problem, the obvious approach is to construct models that capture SSDs’ performance behaviour. However, our previous work has shown the limits of this approach because (a) performance characteristics and energy profiles vary significantly across SSDs, and (b) performance varies in time on a single device based on the history of accesses. The CLyDe project is based on the insight that the strict layering that has been so successful for designing database systems on top of magnetic disks is no longer applicable to SSDs. In other words, our central hypothesis is that the complexity of flash devices cannot be abstracted away as it results in unpredictable and suboptimal performance. We postulate that database system designers need a clear and stable distinction between efficient and inefficient patterns of access to secondary storage, so that they can adapt space allocation strategies, data representation or query processing algorithms. We propose that (i) SSDs should expose this distinction instead of aggressively mitigating the impact of inefficient patterns at the expense of the efficient ones, and (ii) that operating system and database system should explicitly provide mechanisms to ensure that efficient access patterns are favoured.  We thus advocate a co-design of SSD controllers, operating system and database system with appropriate cross-layer optimisations. In this talk, I will report on the lessons we have learnt so far in the project. In particular, I will describe the SSD simulation frameworks that we have developed to explore cross layer designs: EagleTree and LightNVM. I will discuss our findings on the importance of scheduling within an SSD. I will present our contribution to the re-design of the Linux block layer, that makes it possible for Linux to keep up with SSD performance on multi-socket systems. Finally, I will present preliminary results on the co-design of file system and SSDs.   CLyDE is a joint project between IT University of Copenhagen and INRIA Paris Rocquencourt, started in 2012 and funded by the Danish Council for Independent Research.   Bio: Philippe Bonnet is associate professor at IT University of Copenhagen. Philippe is an experimental computer scientist focused on building/tuning systems for performance and energy efficiency. Philippe’s research interests include database tuning, flash-based database systems, secure personal data management, sensor data engineering.