פרופ' מריו נמירובסקי (ברצלונה)
יום שלישי, 26.2.2013, 11:00
חדר 337, בניין טאוב למדעי המחשב
This talk is to motivate an interactive discussion on a couple of new research proposals that presently I have been pushing (DCR and FOG). The markets of embedded real-time devices and datacenters/clouds are destined to converge. The explosive growth of the Internet of Things (IoT) will rapidly push the need to run time-critical applications in the cloud, not only for reducing costs but also for simplifying management and achieving increased availability and efficiency. However, failure to provide guarantees for real-time deterministic execution in the cloud will drive most IoT applications to be implemented locally, hence defeating the advantages of cloud computing. The key idea of Deterministic Cloud Research (DCR) is to leverage on real-time systems and to extend their well-established concepts and practices to datacenter. DCR will facilitate the creation of novel applications, improving predictability. DCR will specify the deterministic API that will be common to both embedded and datacenter environments, thus enabling simple migration from local to a datacenter execution.
However, not all the applications can be offloaded to the cloud, particularly those that require real-time low-latency, aggressive aggregation, reliable connectivity or physical context awareness of the surrounding world. Representative examples of services that cannot be offloaded to today´s cloud: first, services requiring real-time processing cannot depend on the variable latency of the distant cloud and second, applications running in vehicles with poor and intermittent connectivity (e.g., an airliner). As a result, there is a need for a fresh approach; this new concept is FOG. FOG is a highly distributed computing and storage platform at the network from the core to the edge. FOG is not a cannibalization of the cloud but rather an extension that enables new applications that cannot be serviced by the current cloud paradigm.
Mario Nemirovsky is an ICREA Research Professor at the Barcelona Supercomputer Center, where he has been since 2007. He holds 62 patents and has authored over 30 research papers. Mario is a pioneer in multithreaded hardware-based processor architectures. During his tenure with the University of California, Santa Barbara, Mario co-authored some of the seminal works on simultaneous multithreading. Mario has made key contributions to other areas of computer architecture, including high performance, real-time, and network processors. He founded ConSentry Networks, Inc. where he served as CTO and VP Chief Scientist. He was the architect of ConSentry high performance processor (LSP-1) in which he pioneered the concept of Massively Multithreading (MMT). Earlier, Mario founded Flowstorm and XStream Logic, Inc. Before that, he was a chief architect at National Semiconductor, PI Researcher at Apple Computers, and Chief Architect at Weitek Inc. As chief architect at Delco Electronics, General Motors (GM), he architected the GM multithread engine controller. He received his PhD in ECE from University of California, Santa Barbara in 1990.