The world of software is becoming increasingly parallel. This change is visible both in the small scale (more CPUs per machine, more cores per CPU and more threads per core) and in the large scale (large elastic clouds of many virtualized machines). But developing parallel systems is hard! The Meanwhile track covers current and emerging programming models for the multi-core and cloud era.
Recent hardware trends - the slowing of processor frequency scaling in favor of multiple cores, the proliferation of specialized compute devices such as graphics processors, and the emergence of cloud computing - are pushing parallel programming techniques into the mainstream. Recent developments in computer hardware architecture, describe the challenges inherent in parallel programming and introduce tools and libraries that make parallel programming accessible to mainstream developers.
Kerry Hammil is a Senior Program Manager with Microsoft Research. Her current areas of focus include data parallel programming including work on the Microsoft Research Accelerator project. In the past she has contributed to 2D and 3D graphics APIs for various products at Microsoft and she spent a year working at a software startup in the photo industry.
By 2009 almost all servers, laptops and desktop PCs will have multicore processors, the software development community is rather slow at adopting the parallel paradigm. In this presentation we give some tips that will help the developer avoid some of the known pitfalls in parallel programming. These tips are based on experience gained in real-world applications. The underlying technical limitations and challenges of programming for multicore are discussed along with implementation examples
Bernth is a Technical Consulting Engineer at Intel, which he joined 1982 as a Training Specialist in Processor Architecture, Software and RTOS. During the last 10 years he has been working in the Intel Compiler Lab and is a regular speaker at technical conferences in Europe. Prior to Intel, Bernth worked 5 years with Ericsson as a Software Instructor.
Stuart Dabbs Halloway is a co-founder of Relevance, Inc. Stuart is the author of Programming Clojure, Component Development for the Java Platform and Rails for Java Developers. Stuart regularly speaks at industry events including the No Fluff, Just Stuff Java Symposiums, the Pragmatic Studio, RubyConf, and RailsConf.
In this talk, I will describe the principles behind Erlang-style Concurrency - what problems it was designed to solve, and how it fundamentally changes the way you go about structuring your programs. I will illustrate how to achieve great scalability on multicore and in compute clouds, without sacrificing clarity or your own sanity.
Ulf Wiger has used Erlang since 1993, as one of its very first commercial users. At Ericsson, he was Chief Designer for the AXD 301 project - possibly the most complex system ever built using a functional language, and famous for its remarkable in-service performance. Since February this year, Ulf is the CTO of Erlang Training and Consulting Ltd.
Hadoop is an open source implementation of Google's Map Reduce and
Google File System (GFS), a distributed file system and processing
engine. Hadoop is used in industry to store and analyze vast amounts
of data on hundreds or thousands of commodity servers. This seminar will talk about Hadoop's uses and implementation from a high level, and introduce some of the other projects and tools that exist in the Hadoop ecosystem that can help companies and individuals with their big data problems.
Alex Loddengaard is part QA engineer, part operations engineer, part
support engineer, and part Hadoop trainer at Cloudera. He spends most of his time
deploying and testing Hadoop. He has also contributed to the open-source Hadoop
project itself. While at the University of Washington, Alex was awarded the Bob Bandes Memorial Award for Excellence
in Teaching his first year as a teacher's assistant (TA), and later grew to become a guest lecturer and head teacher's assistant.