The First Workshop on Benchmarks, Performance Optimization, and Emerging hardware of Big Data Systems and Applications (BPOE 2013) In conjunction with 2013 IEEE International Conference on Big Data (IEEE Big Data 2013) October 8, 2013, Silicon Valley, CA, USA
July 30, 2013: Due date for full workshop papers submission
August 30, 2013: Notification date
September 10, 2013: Camera-ready deadline of accepted papers
October 8, 2013: Workshops
Big Data has emerged as a strategic property of nations and organizations. They are driving needs to generate values from Big Data. However, the sheer volume of big data problems requires significant storage capacity, transfer bandwidth and computing powers. It is expected that more and more research questions can only be fulfilled by systems with unprecedented scales, due to the daunting volume, velocity and variety of Big Data. Facing with different Big Systems, owners of Big Data can hardly make a choice on which system is the best suited for their specific requirement. The users of Big Data also face the challenge on how to optimize the system and their solutions to get the most for existing data collection. On the other hand, system researchers and developers are working onnew hardware architecture, new system models, data management and mining techniques to optimize the performance in dealing with Big Data. To lay the ground for development of standards for measuring the efficiency of hardware and software technologies dealing with Big Data is extremely urgent. This workshop engages industry and academic experts to address the above problems. The workshop is organized around practical big data application examples, novel hardware and software applications, and the state-of-the-art techniques on benchmarking, characterizing, and optimizing Big Data systems.
Benchmarks provide fair basis for comparison among different Big Data systems. Besides, benchmarks represent typical needs of system support from Big Data applications. Together with workload characterization of typical Big Data applications, benchmarking results can thus enable active improvements of Big Data systems. Further, performance analysis and optimization of overall systems and applications can help target at a specific Big Data problem better. As full-scale experiments are not always possible and costly for Big Data, new experimental methodologies are needed to synthesize PB-scale Big Data and simulate 100K-node system, so as to obtain insights of large-scale systems from small-scale deployments. Characteristics of Big Data Systems bring unique challenges for system benchmarking, optimizing, and workload characterization. The system complexity and application diversity make it difficult to develop workloads and design use cases, and the rapid development of new Big Data Systems re quires benchmarks to keep in step with improvements of underlying systems. In addition, the huge volume of Big Data poses big infrastructure challenges for performance optim ization, including data workloads at a Petabyte scale, massively parallel data processi ng, and system configuration tuning in data centers and cloud platforms. All these aspects are interrelated and challenging, and they can also lead to great opportunities for designi ng new innovative Big Data infrastructure.
Additionally, new computing and storage technologies are also in rapid development and pushing for new high-end hardware and new software models. For examples, GPGPU and new Intel Xeon Phi processors can easily incr ease the number of processing cores within one box; new solid state drive and flash drive can provide orders of magnitude more IOps than traditional hard drives. These new hardware bring new opportunities for performance improvement but also new challenges. For exam ple, the overall performance bottleneck of a problem can be shifted, requiring different workload balancing strategies due to significant performance boost of a particular hardware. Those new technologies also need to be introduced to the research community with t horough evaluation to identify their strengths, limitations and further optimization.
This unique combination of opportunities and challenges attracts much attention from both academia and industry. The BPOE work shop aims at bringing researchers and practitioners in related areas together to discuss the research issues at the intersection of these areas, and also to draw much atte ntion from the general architecture, systems, data management and mining research communities to this new and highly promising field.