COMPUTER MEASUREMENT GROUP - India Hyderabad Chapter Strategies to Optimize Cloud Costs By Cloud Performance Monitoring October 2013 www.cmgindia.org Computer Measurement Group, India 1
About Me Credentials 15 Years of experience Enterprise Architect and Cloud Evangelist Open CA Level 2 Certified Specializations Architecture and Design of Large Scalable Systems Performance Tuning Microsoft Platform Current Role Technology Excellence Group Head HiTech Industry Solutions Unit@TCS My Role in CMG India Regional Facilitator CMG India Hyderabad Chapter
Key Take Aways Understanding the Cloud Economics Myths and Truths of Cloud Tenets of Designing for Cost in Cloud Cloud Cost Variables Computer Measurement Group, India 3
Key Take Aways Understanding the Cloud Economics Myths of Cloud Tenets of Designing for Cost in Cloud Cloud Cost Variables Computer Measurement Group, India 4
Cloud Economics Cost Usage Capacity Capacity is CapEx and cyclically driven Usage follows predictable pattern Cost increases with time and complexity Capacity scales up and down with demand Usage follows predictable pattern Cost matches capacity Pay as you consume Capacity is provisioned based on expected Load Usage follows predictable pattern Cost goes up and to the right within a moment Computer Measurement Group, India 5
Efficient Computing Strategy Measure everything which you can control and mange Continuous monitoring Assess meter cost of variables Capacity Usage Know Your measuring Units Computer Power CPU/hour Memory usage - GB/ hour Storage Consumption GB/stored Network utilization GB/managed Follow Cyclic process to inform, improve and maintain. Computer Measurement Group, India 6
Cloud Myths and Truth Cloud is Cheap More Expensive Life for Like It is different in terms of Scalability Resilience Redundancy Flexibility More of OpEx In reality it is more costeffective when there is large workloads
Cloud Myths and Truth Cloud is Cheaper than Do It Yourself Cloud is cheaper Vs Fully loaded server cost Do we understand the Infrastructure true Costs? Cloud sometimes looks expensive but it is not Both Netflix and Zynga Cloud models are right in their terms http://www.scalevp.com/how-zynga-andnetflix-use-the-cloud
Cloud Myths and Truth Pay for What you Use Pay for What you Provision
Cloud Myths and Truth One need to understand Cost Dynamics I can Migrate easily Data IN is free Data OUT is charged Thumb Rule : Data attracts more data More cost is involved to move more data
Cloud Myths and Truth Cloud Providers manage their Cloud is Managed cloud and NOT your applications You still need to take care of your operations
Cloud Myths and Truth Cloud Waste is because inefficient Cloud is more Efficient design practices and it is not because of technology
Tenets of Designing for Cost on AWS Minimize the Always ON and Use Reserved Instances Design for SPOT Store it RIGHT and leave it in Cloud Leverage Scalable and On demand services Scale IN Automatically Clean up After yourself
Tenets: Minimize the Always ON and use Reserved Instances Key Principles Elasticity Optimize based on real-time demand
Tenets : Design for SPOT Key Principles SPOT instances can give significant savings over on-demand instances Build architecture for elasticity Use Case Batch Processing Hadoop Scientific Computing Video and Image Processing Testing HPC Financial Types of Applications Generic background processing Hadoop / Map Reduce Processing type jobs Scientific trials/ simulation jobs / analysis in Physics, chemistry Transform videos in different specific formats Provide testing of software, web sites etc. Parallel Jobs Trading, Hedgefund analytics
Tenets : Store it RIGHT and leave it in Cloud Key Principles Pick the right storage model Minimize the in and out of data For big data workloads, storage and data transfer influence cost. Service Storage Cost Durability Access Example S3 $0.14 / GB Very High Whole object Source videos S3 RRS $0.093 / GB High Whole object Transform videos EBS $0.10 Medium Block Device Database Partition Ephemeral Storage CloudFront Download CloudFront Streaming Free Low Block Device Logs Variable N/A Whole object S/W downloads S3 Cost N/A Partial Object Static videos Live Streaming EC2 + Storage N/A Partial Object Live Video
Cost Variables
Cost Variables
Right Sizing helps to Optimize Cost Examples Development Environment : 40 hours / Week (24% availability) Load Test : 8 Hours / day - Once off Back-Office Applications : 40 hours/ week heavy; Outside hours 25% capacity Travel Sites / E-commerce portals : Consumer Centric; Market-driven; Seasonal Computer Measurement Group, India 19
References 1. AWS Economics Centre : http://aws.amazon.com/economics/ 2. Cloudonomics Book: http://www.amazon.com/cloudonomics-website-business- Value-Computing/dp/1118229967 3. Microsoft Cloud Economics: http://www.microsoft.com/enus/news/presskits/cloud/docs/the-economics-of-the-cloud.pdf 4. AWS Costs Cheat Sheet: https://blog.cloudvertical.com/2012/10/aws-cost-cheatsheet-2/ 5. On-Premise vs. Cloud Cost Model: http://andrewmcafee.org/2012/10/mcafeecloud-costs-google-model 6. Data Gravity: http://datagravity.org/ 7. AWS Cost Cheat Sheet Computer Measurement Group, India 20
Thank You