amazon web services - cs.ucf.edudcm/teaching/cda5532-cloudcomputing... · amazon web services:...

Post on 05-Sep-2020

2 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Amazon Web Services:Caching and Storage Options

Summarized by: Michael Riera

9/17/2011

University of Central Florida – CDA5532

Agenda

• Purpose

– Cost Saving and reduced latency

• Cache Framework and Relative Options

• Related Work• Related Work

• Experiment Results

• Conclusion

Introduction

• Evaluating the performance and costsassociated with a number of caching andstorage options offered by the Cloud.

– High performance system:– High performance system:

• Benefits: fetching, data fetching and reuse

– Trade-off between performance and usage cost

AWS Options

• Approach 1: Instantiating VM and save data inmemory or in storage (limited capacity)– Higher Cost, fastest access time

• Approach 2: Persistent storage with AWS S3• Approach 2: Persistent storage with AWS S3– Overhead in slower throughput

• Approach 3: Persistent storage with ElasticBlock Storage– Storage can only connect to 1 instance at a time

– EC2 can connect to multiple EBS

Cache Framework• Automatically allocatesmachine instances as toexpand cache’s capacity

• Managed using aconsistent hashing approach

• Data set k

• Nodes which should store• Nodes which should storedata is referenced by thenearest bucket larger than k

• Each node stores a B+-Tree which indexes the storeddata.

Experiment Setup

• Resource Configuration: We have run experiments over thefollowing configurations:– S3: Data stored as files directly onto the S3 storage service

(persistent).– ec2-m1.small-mem: Data stored in memory on Small EC2 instance

(volatile, moderate I/O).– ec2-m1.small-disk: Data stored as files on disk on Small EC2 instance– ec2-m1.small-disk: Data stored as files on disk on Small EC2 instance

(volatile, moderate I/O).– ec2-m1.small-ebs: Data stored as files on a mounted Elastic Block

Store volume on small EC2 instance (persistent, moderate I/O).– ec2-m1.xlarge-mem: Data stored in memory on Extra Large EC2

instance (volatile, high I/O).– ec2-m1.xlarge-disk: Data stored as files on disk on Extra Large EC2

instance (volatile, high I/O).– ec2-m1.xlarge-ebs: Data stored as files on a mounted Elastic Block

Store volume on Extra Large EC2 instance (persistent, high I/O).

Experiment Application

• Land Elevation Change Process

– Input: A Location L, and a coordinate t, where t’ :t< t’ representing the times of interest

– Output: Locates Digital Elevation Models (DEM)– Output: Locates Digital Elevation Models (DEM)Files with respect to (L, t) and (L, t’) representinglarge matrix.

• 2000 Random queries with different instanttypes

Experiment Results

Experiment Results

Experiment Results

top related