Today’s interview features Jason Dai, Engineering Director and Principal Engineer for Intel SSG (Software and Services Group). Jason will be co-presenting How to Weigh an Elephant, a discussion on benchmarking big data systems. Jason’s co-presenter will be Vin Sharma, a software strategist at Intel responsible for planning and marketing Intel contributions to open source datacenter software projects including Apache Hadoop, OpenStack, KVM and enterprise Linux.
Tell us about your presentation.
Jason: The focus is on benchmarking. As any good engineer knows: “if you cannot measure it, you cannot improve it”. Benchmarking is the quantitative foundation of any computer system design and research and that’s especially true in the world of Apache Hadoop and big data. This presentation will cover our recent work on HiBench, an open source Hadoop benchmark suite widely used by Hadoop users. It will also cover different existing benchmark approaches for Hadoop such as trace-based benchmarks (e.g. GidMix3) so that attendees can understand their tradeoffs. There will also be a discussion on fostering a community for Hadoop benchmarking.
What do you expect will be the single biggest takeaway of your presentation for attendees?
Jason: Realistically benchmarking Hadoop is rocket science. Good benchmarks (e.g. HiBench) and community-driven efforts are critical for Hadoop benchmarking.
Tell us about your current role and how you interact with Hadoop.
Jason: I’m currently an Engineering Director and Principal Engineer in Intel SSG (Software and Services Group), leading the open source Hadoop & Big Data engineering team. Our team is working on many open source projects related to Hadoop, such as HiBench (Hadoop benchmark suite), HiTune (Hadoop performance analyzer), and Project Panthera.
What are you most looking forward to at Hadoop Summit?
Jason: Meeting not only the developers but also the users of Hadoop, so as to understand their challenges that may or may not be served by Hadoop today.
What other presentations are you most looking forward to attending?
Jason: Presentations on new projects within the Hadoop ecosystem, such as Drill, BookKeeper, etc.