OpenPOWER Foundation EU Summit 2018

Tackling the Challenges of Massive Data Analytics
In the session, we'll take a look at how combining modular components and sophisticated algorithms can deliver fast analytics performance, by limiting the effect of I/O on data-intense queries and models. We'll show how arranging data for the GPU, combined with fast GPU compression, metadata mapping, and several other techniques can accelerate real database physical operators. We will look at how the combination of high throughput processors like IBM POWER9 and SQream's GPU database delivers an analytics platform that breaks down these barriers. It will show how the increased performance and processing power available in POWER9 combined with SQream's advanced algorithms for fast data access, and ability to JOIN any table on any keys without manual indexing or pre-aggregating, enables organizations to query much larger data stores, much faster. We will also look at how high throughput compute devices, like IBM POWER9 and NVIDIA Tesla GPU accelerators can handle data significantly faster than other data solutions. This ability, coupled with improved I/O through clever software techniques, can not only reduce latency for the AI/ML/DL pipeline, but also operate at much larger scales. High throughput at-scale access to raw data increases the accuracy of data science models and improves predictions, by training and predicting on full data sets, rather than cut-down versions.