|Name:||Re-Imagining Storage for HPC & Big Data|
|Time:||Thursday, June 26, 2014
09:00 am - 09:20 am
CCL - Congress Center Leipzig
|Breaks:||07:30 am - 10:30 am Welcome Coffee|
|Speaker:||Bill Moore, DSSD|
|Abstract:||In today’s data-driven economy, corporations have access to massive amounts of unstructured data and wish to analyze and act upon this information to deliver better products and services to their customers. Applications ranging from real-time fraud analysis to facial recognition and bioinformatics are requiring a new data-centric storage architecture that can store PBs of data and access them from many CPUs at very low latency. This shift requires a rethinking of the storage architecture.
At the same time the underlying technology for these storage systems is beginning to move to Flash technology that can dramatically accelerate these applications while reducing power. The question is what’s the best way to deliver the full performance of Flash to HPC and Big Data applications, efficiently and reliably. Existing Flash-based solutions fall into two categories: First, there are direct-attached SSDs that deliver high performance but are dedicated to each server. And second, there are Flash arrays that deliver better reliability and scalability in a shared platform but rely on legacy storage fabrics that limit performance.
This talk will discuss pro’s and con’s of existing storage solutions and propose a re-imagined storage architecture that is optimized for HPC and Big Data.