Apache Spark Acceleration Using Hardware Resources in the Cloud, Seamlessly - Databricks

Apache Spark Acceleration Using Hardware Resources in the Cloud, Seamlessly

Download Slides

InAccel provides high performance accelerators for your application based on novel hardware reconfigurable engines as IP blocks. The hardware accelerators can be deployed in the cloud, like Amazon AWS, using the f1 accelerators and can be integrated to widely used frameworks like Spark and PostgreSQL. The main novelty is that the users do not need to change the original code as the accelerators are deployed as software packages.

In this talk we will show how machine learning applications (e.g. logistic regression and K-means), based on Spark, can be accelerated by 3x-10x using hardware accelerators that are deployed in the Amazon AWS using f1 accelerators without any changes on the Spark code. InAccel provides all the required APIs for the integration on Spark using Java, Python or Scala. The utilization of hardware accelerators can also be used to reduce the OpEx as less resources and less time is required for the processing of the data.

Session hashtag: #HWCSAIS16

About Christoforos Kachris

Christoforos Kachris is the founder and CEO of InAccel technologies that develop hardware IP cores for the acceleration of machine learning applications over Spark and other frameworks. InAccel provides modules for Amazon EC f1 platform that allows the speedup of Spark applications. His main expertise in on hardware acceleration of applications like network processing, image processing, high performance and cloud computing. he is the Technical Project Manager of the H2020 VINEYARD project working on efficient utilization of reconfigurable accelerator-based servers in the data centers