Sabri SKHIRI

Chief Architect of the R&D Architecture, Huawei EU Research Center

Sabri leads the R&D architecture team (Carrier Software BG) of the European Research Centre in Belgium. His team leads researches on new Service Delivery Platform models, Cloud computing & Big Data PaaS, high performance messaging and integration technologies, Business and User intelligence, distributed & in-memory machine learning platform, distributed social applications, distributed Pattern Matching engine and Stream Processing, and finally Model-Driven approaches for IDEs. Sabri is also at the head of the R&D department of a Belgian Research Company specialized in Distributed computing & Machine learning.
Sabri started his career as a researcher at the Université Libre de Bruxelles (ULB) in the graph layout area. He published several scientific papers on generic algorithms which draw graphs according to the semantics of the application domain. He then joined Alcatel-Lucent R&D where he was responsible for the evolution of the OSP Telecom Application Server towards JAVA & JavaEE. His responsibilities extended to the innovation projects and to the development of service prototypes for customers.
In 2009, Sabri founded EURA NOVA, a research company, with two partners. Since, his role has been to create the bridges between innovation, research and customer challenges in various sector, such as Telecom, Pharmaceutical, Finance and Banking industries. In this scope, Sabri created significant relationships with EU Academic Research Centers in Belgium, France, Germany and Spain. He is also a committer on open source projects such as roq-messaging, Steffi GraphDB and AROM processing.
Sabri also published several papers in different International Scientific conferences in domain like distributed computing, graph mining, conceptual modeling and elastic architectures.

演讲主题：Lambda Architecture 2.0 Convergence between Real-Time Analytics, Context-awareness and Online Learning

最近实时分析方面的商业案例已经开始兴起了。这些商业案例需要分析实时数据的能力，包括： 1. 检测出感兴趣的情况； 2. 利用在批量分析中计算出的预测模式，来作出最优决策； 3. 用最新的数据来更新预测模型，使模型尽可能贴近不断变化的现实。此外，在电信行业中，现在实时数据的负荷和吞吐量已经非常高，可以达到每秒几百万事件。
为了满足这些实时分析的能力并且提供足够的性能，采用新架构模式和新的技术是必需的。在这次演讲中，我将从知名的LAMBDA架构开始，介绍那些新架构模式。我还将介绍我们是如何改进性能不够好的传统复杂事件处理技术，改造成完全分布式的、可扩展的模式匹配引擎。最后，我将谈到新的为高性能在线机器学习而设计的内存计算模式。
Recently new business cases have been emerging in Real-Time analytics. Those cases required the ability to analyze the data in real-time: (1) detecting interesting situation, (2) leveraging predictive models calculated in batch to take an optimal decision and even (3) updating the predictive models with latest data in order to let the models sticking as much as possible to the evolving reality. In addition, in the telecommunication sector, the data workload and throughput are significant and can reach several millions of events per second.
In order to achieve those requirements with those performance constraints, new kind of architecture patterns and new technologies are mandatory. In this talk, I will describe one of those new patterns starting from the well-known Lambda architecture. I will also introduce how we could improve the traditional and not high performance Complex Event Processing Technology to a fully distributed and scalable Pattern Matching Engine. Finally, I will speak about new kind of in-memory processing designed only for High Performance On-Line Machine Learning.