Transcript pptx
Distributed Systems Foundations Lecture 0 Evolution of computing history • Main Frame with terminals • Network of PCs & Workstations. • Client-Server • Now, moving forward to Large cloud. CS 271 2 Cloud Reality: Data Centers CS 271 3 Cloud Computing: Why Now? • Experience with very large datacenters – Unprecedented economies of scale – Transfer of risk • Technology factors – Pervasive broadband Internet – Maturity in Virtualization Technology • Business factors – Minimal capital expenditure – Pay-as-you-go billing model CS 271 4 Economics of Cloud Users • Pay by use instead of provisioning for peak Resources Resources Capacity Demand Capacity Demand Time Time Static data center Data center in the cloud Unused resources CS 271 Slide Credits: Berkeley RAD Lab 5 Economics of Cloud Users • Risk of over-provisioning: underutilization Capacity Resources Unused resources Demand Time Static data center CS 271 Slide Credits: Berkeley RAD Lab 6 Economics of Cloud Users Resources Resources • Heavy penalty for under-provisioning Capacity Demand Capacity 2 1 3 Time (days) 2 1 Lost revenue 3 Demand Resources Time (days) Capacity Demand 2 1 3 Time (days) Slide Credits: Berkeley RAD Lab CS 271 Lost users 7 Cloud Properties • Commodity hardware • Large Scale • Elasticity CS 271 8 Elasticity in the Cloud Client Site Client Site Client Site Load Balancer (Proxy) App Server App Server App Server CS 271 App Server App Server 9 Why does this work? • As long as requests are stateless, we can add more resources, thus providing: •Scale •Elasticity CS 271 10 But, most services need DATA! • Challenges: –How to scale with the increasing amounts of data –Where to store the data –Accessing data on multiple sites –Failures CS 271 11 Need • Fault-tolerance: – Replication • Large scale data: – Partition data across multiple servers • Multiple Servers: – Time: clocks – Coordination: mutual exclusion, leader election – Consensus: Byzatine agreement, Paxos, etc – Distributed state – P2P CS 271 12 Explosive Data Growth • Wikipedia has over 3.5 million pages. • Flickr members uploaded over 5 billion photos • You Tube has 35 hours of videos uploaded each minute • “more video uploaded to YouTube in the past two months than there would have been if ABC, CBS, and NBC had been airing 24/7 since 1948!” Gartner 2010 CS 271 13 Twitter 6th B’day: March 21, 2012 • First tweet: "inviting coworkers“, 2006 • a record of 6,939 tweets per second immediately after Japan quake on March 11 2011, and 177 million tweets rest of the day. • It took three years, two months and a day for Twitter to get to one billion tweets. • Twitter now averages 140 million tweets a day. CS 271 14