Description: With the rise of mobile, search and social media users are now conditioned to expect instantaneous results via intuitive drag and drop interfaces. They bring these expectations to every part of their life especially their work. No longer is it acceptable, from a user’s point of view, to wait for an answer to a question. The fact that their question may need to churn through petabytes of data residing on hundreds of servers is irrelevant to the end user. They want their data and they want it now!
Abstract: Managing users’ expectations is hard on any platform, but on Hadoop it is nearly impossible. In most large businesses, users are migrating to Hadoop from large enterprise systems that have had thirty plus years to be tuned and optimized for the users’ workload.
The current big data challenge is the mixture of high expectations, ever increasing critical workloads and massive amounts of data. Delivering large amounts of data reliably is no longer good enough. Big data users are now demanding speed.
LLAP is one of the new products promising to meet these user expectations by delivering Big Data in real time.
In this presentation, Walmart Labs will detail our LLAP proof of concept and present whether our users’ expectations were met. Areas covered in this presentation:
• Current Hadoop SQL performance
• LLAP: low latency analytical processing or live long and Process?
- What is LLAP?
- LLAP installation
- LLAP tuning
- How does it compare to traditional Hive and other Hadoop accelerators?
• Key learnings