17
Title Demonstrating Benefits of Hyp Accelerati Roop Ganguly, Solution Architect

Demonstrating the Benefits of Hyper-Acceleration

Embed Size (px)

Citation preview

Page 1: Demonstrating the Benefits of Hyper-Acceleration

TitleDemonstrat ing the Benefits o f Hyper-Acce lerat ionRoop Ganguly, Solution Architect

Page 2: Demonstrating the Benefits of Hyper-Acceleration

The End of Moore’s Law

350 nm180 nm

130 nm

90 nm

65 nm

1.0

2.0

3.0

1970 1980 1990 2000

Power Wall

GHz

Gordon Moore

Page 3: Demonstrating the Benefits of Hyper-Acceleration

Implications for Big Data

Security AnalyticsRisk Management

Behavioral Analytics

Natural Language Processing

AI/Deep Learning

Machine Learning

Page 4: Demonstrating the Benefits of Hyper-Acceleration

CPU-Bound Applications – A New Bottleneck

40Gb-100GbNetwor

k

Now that faster networking and disk technologies have emerged, CPUs act like “stop signs” for computation

Node 1

Node 2

Node 3

Page 5: Demonstrating the Benefits of Hyper-Acceleration

AcceleratorsMicroprocessor and Cloud Vendors Respond

ASIC

GPU

FPGA

Page 6: Demonstrating the Benefits of Hyper-Acceleration

Data Scientists & Developers

Performance Team

Inhibitor: Programming Model Gapfor Hardware Accelerators

Two wildly different skill

sets

CPU GPU FPGA

Data Science Programming Model

BIG DATA PLATFORMS

Acceleration Programming Model

Programming Model Gap

Page 7: Demonstrating the Benefits of Hyper-Acceleration

Cross Platform

Cross Hardware

Intelligent, automatic computation routing

Zero code change

Introducing BigstreamHyper-acceleration Layer

Dataflow Adaptation Layer

Bigstream Dataflow

Bigstream Hypervisor

HYPER-ACCELERATION LAYER

BIG DATA PLATFORMS

CPU GPU FPGA3X to 30X acceleration

Page 8: Demonstrating the Benefits of Hyper-Acceleration

Accelerated Spark Architecture with Bigstream

Page 9: Demonstrating the Benefits of Hyper-Acceleration

9 9

Business Intelligence Use Case

Page 10: Demonstrating the Benefits of Hyper-Acceleration

Business Intelligence Query

•Based on Transaction Processing Performance Council – Decision Support (TPC-DS) Benchmark

•Spark/SQL Query: SELECT i_item_id , avg(ss_quantity) agg1, avg(ss_list_price) agg2, avg(ss_coupon_amt) agg3 | FROM store_sales, customer_demographics, date_dim, item, promotion WHERE ss_sold_date_sk = d_date_sk AND ss_item_sk = i_item_sk AND…….•Input: approximately 2GB of avro table data •Simultaneously run software-accelerated and unaccelerated on identical Amazon EMR clusters

Page 11: Demonstrating the Benefits of Hyper-Acceleration

Business Intelligence Use Case Demo

Page 12: Demonstrating the Benefits of Hyper-Acceleration

12 12

ETL Adtech Use Case

Page 13: Demonstrating the Benefits of Hyper-Acceleration

Adtech ETL/ML Data Pipeline

Spark Streamin

g

Spark Streamin

g

APPLICATION/WEB

SERVERS KAFKA

clicks

clicks, likes

impressions

USERS

Spark ML

RTB System

s

Distributed messaging system

(tens of servers)

Distributed computation system

(hundreds of servers)

Millions of users

Page 14: Demonstrating the Benefits of Hyper-Acceleration

ETL Use Case Demo

Page 15: Demonstrating the Benefits of Hyper-Acceleration

Announcement –Bigstream onAWS EMR

Page 16: Demonstrating the Benefits of Hyper-Acceleration

Setting the bootstrap script

Bigstream ON EMRAdd the Bigstream bootstrap URLand your cluster has hyper-acceleration

Page 17: Demonstrating the Benefits of Hyper-Acceleration

Thank You