Cloud Computing and Distributed Systems Introduction PDF

Cloud Computing and Distributed Systems Introduction Raja Appuswamy Eurecom Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 1 / 55 Cloud computing: The disruption “In...

Cloud Computing and Distributed Systems Introduction Raja Appuswamy Eurecom Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 1 / 55 Cloud computing: The disruption “In 2020, the global cloud computing market was valued at $371.4 billion, and it is estimated that by 2025 it will rise to a staggering $832.1 billion.” – Marketsandmarkets “74% of Amazon’s operating profit comes from AWS” – Amazon “80% of organizations will migrate toward the cloud by 2025.” – Gartner “50% of all data will be held in the cloud by 2020. Cloud data centers will process 94% of workloads in 2021.” – IDC & Cisco “Global data centers used roughly 416 terawatts (3% of the total electricity) last year, nearly 40% more than the entire United Kingdom.” - Forbes “Big data solutions via cloud subscriptions will increase about 7.5 times faster than on-premise options.” - Forrester “AI without the cloud is tough” – Information Age Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 2 / 55  Course Overview This Course What you will learn (roadmap) Economic foundations Service models Infrastructure foundations Virtualization, containerization, serverless functions Systems foundations Hadoop, Apache Spark Relational databases, Distributed file systems, maybe bitcoin & blockchain Programming foundations Map—reduce and functional programming SQL and NoSQL Algorithmic foundations Consistency, Serializability, Transactions Atomic commitment, two-phase, three-phase commit Consensus, PAXOS, CAP theorem Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 3 / 55  Course Overview Who is this course for? Cloud developers, architects, data engineers, data analysts, … Cloud computing sector is expected to grow 14 percent annually and create one million new jobs in 2022 Right now there are an estimated 5.6 million cloud- related jobs worldwide $80,000 to $200,000 per year You will ace system design interview with this course You will be on your way to getting certified Requirements Some familiarity with operating systems concepts Some familiarity with computer architecture Knowledge of python, git, web frameworks Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 4 / 55  Course Overview Grading Final Exam: 60% of the overall grade Entirely based on course lectures Design questions, problems, … Labs: 40% of grade Assisted labs on GCP 50% discount on Associate Cloud Engineer certification DIY labs on Azure Maybe an assisted lab on Spark (being worked out) Distributed systems lab (or) a customized research project (prereq: strong C++ bg.) Bonus component In class quizzes Based on previous lecture, after mid-class break Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 5 / 55  Course Overview How to make the most of this course? Attend classes Many discussions in live classes Lecture notes in Moodle Past lectures on mediaserver PLAN AHEAD FOR LABS & PROJECT Labs/project will be hard if you have non-CS background Everything can be done remotely Books Designing data intensive applications Spark: The Definitive Guide Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 6 / 55  Introduction to Cloud Computing Introduction to the Cloud Computing Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 8 / 55  Introduction to Cloud Computing We live in a world of data Figure: Data deluge. Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 9 / 55  Introduction to Cloud Computing Big Data Big data is defined as large pools of data that can be captured, communicated, aggregated, stored, and analyzed. Data continues to grow Figure: Global datasphere Applications are becoming data intensive More data leads to better accuracy With more data, accuracy of different algorithms converges Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 10 /55  Introduction to Cloud Computing Let’s look at your data. You want to access, shared, process your data from all your devices, anytime, anywhere. Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 11 /55  Introduction to Cloud Computing How will we manage all this data? Manage it ourselves? How do we store it? How do we share it? How can we enable access to it from any place? How do we process all of it? How do we secure it?.... What if it is managed by someone else? Someone provides a management “service” You pay a subscription for this “service” Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 12 /55  Introduction to Cloud Computing Utility–Product–Service lifecycle: Water Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 13 /55  Introduction to Cloud Computing Utility–Product–Service lifecycle: Electricity Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 14 /55  Introduction to Cloud Computing Generalizing the lifecycle Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 15 /55  Introduction to Cloud Computing Cloud computing: The prophecy In 1965, MIT's Fernando Corbató and the other designers of the Multics operating system envisioned a computer facility operating “like a power company or water company”. Plug your thin client into the computing Utility and Play your favorite Intensive Compute & Communicate Application Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 16 /55  Introduction to Cloud Computing Cloud Computing Transformation of IT from a product to a service Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 17 /55  Introduction to Cloud Computing Formal definition Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 18 /55  Cloud infrastructure IT as a service How do we offer IT as a service? Different users have different needs Average end user Mobile app developer Enterprise systems architect Let us look at some service models Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 19 /55  Cloud infrastructure Basic cloud service models Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 20 /55  Cloud infrastructure SaaS Software is delivered as a service over the Internet, eliminating the need to install and run the application on the customer’s own computer Simplifies maintenance and support You use SaaS products everyday Gmail, Google docs, Youtube,... Salesforce.com is a popular commercial pioneer (ERP, CRM,...) Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 21 /55  Cloud infrastructure PaaS The Cloud provider exposes a set of tools (a platform) and APIs which allows users to create SaaS applications The SaaS application runs on the provider’s infrastructure The cloud provider manages the underlying hardware and requirements Examples: Google App Engine, Windows Azure Web App service Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 22 /55  Cloud infrastructure IaaS The cloud provider leases to users Virtual Machine Instances (i.e., computer infrastructure) using the virtualization technology The user has access to a standard Operating System environment and can install and configure all the layers above it Ex: AWS EC2, Rackspace, Google Compute Engine Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 23 /55  Cloud infrastructure Other services models Hardware-as-a-service (HaaS) You get access to barebones hardware machines, do whatever you want with them, Ex: Your own cluster https://www.youtube.com/watch?v=pqfd4t9ISHY X-as-a-service, where X can be Backend (BaaS), Desktop (DaaS),... Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 24 /55  Cloud infrastructure Cloud Computing Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 25 /55  Cloud infrastructure Cloud Infrastructure Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 26 /55  Cloud infrastructure What is a server? Servers are computers that provide “services” to “clients” Typically designed for reliability and to service a large number of requests Dual-socket servers are the fundamental building block of cloud infrastructure Organizations typically require many physical servers to provide various services Web server, database server, mail server,... Server hardware is becoming more compact conserving floor space improving manageability power and cooling Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 27 /55  Cloud infrastructure What is a rack? Servers are grouped, placed, and organized in racks Equipment are designed in a modular fashion to fit into rack units (1RU = 4.45cm) A single rack (6 ft or 180cms) can hold up to 42 1U servers Figure: Global datasphere Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 28 /55  Cloud infrastructure What is a data center? Facility used to house a large number of computer systems and associated components Air conditioning Power supply Hazard protection Security and monitoring systems Networking and connectivity Let’s take a look at two datacenters https://www.youtube.com/watch?v=zDAYZU4A3w0&t=416s https://www.youtube.com/watch?v=L2oJw1a_qEM Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 29 /55  Cloud infrastructure Problems with privately owned data centers Expensive to setup (High capital expenses or CAPEX) Real estate, server and peripherals,... Expensive to operate (High operational expenses or OPEX) Energy costs (Good data centers have efficiency of 1.7, 0.7 Watts lost for each 1W delivered to the servers) Administration costs Difficult for applications to grow/shrink How do we map applications to servers? What if we over/under provision? Low utilization (30% server usage considered good) Throw money at the performance problem (peak provisioning) Uneven application fit: each server has CPU, memory, and disk: most applications exhaust one resource, stranding the others Uncertainty in demand: Demand for a new service can spike quickly Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 31 /55  Cloud infrastructure What if Turn the servers into a single large resource pool and let services dynamically expand and contract their footprint as needed? Two main requirements: Means for rapidly and dynamically satisfying application fluctuating resource needs Provided by virtualization Means for servers to quickly and reliably access shared and persistent data Provided by programming models and distributed file/storage/database systems Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 32 /55  Cloud infrastructure What is a cloud then? Single-site cloud A data center hardware and software that the vendors use to offer the computing resources and services Geographically distributed cloud Multiple such sites, with each site perhaps having different structure and services https://www.youtube.com/watch?v=47e_3WBCe-Q Figure: Azure: 1 million servers, 100 data centers across 90 countries. Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 33 /55  Cloud infrastructure Cloud h/w-s/w stack Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 34 /55  Cloud infrastructure The Cloud Stack Applications Cloud applications can range from Web applications to scientific computational jobs Data Old SQL systems (Oracle, SQLServer) NoSQL systems (MongoDB, Cassandra) NewSQL systems (TimesTen, Impala, Hekaton) Runtime environment Runtime platforms to support cloud programming models Example: Hadoop, Spark Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 35 /55  Cloud infrastructure The Cloud Stack Middleware Platforms for Resource Management, Monitoring, Provisioning, Identity Management and Security Operating systems Standard Operating Systems used in Personal Computing Packaged with libraries and software for quick deployment and provisioning E.g., Amazon Machine Images (AMI) contain OS as well as required software packages as a “snapshot” for instant deployment Virtualization (serverse, storage, networking) Key enabler of cloud computing Providers resource virtualization, multitenancy Ex: Amazon EC2 is based on the Xen virtualization platform, Azure based on HyperV Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 36 /55  Cloud infrastructure Cloud service models and the cloud stack Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 37 /55  Cloud infrastructure Cloud Computing Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 38 /55 “A Cloudy History of Time” First large datacenters: ENIAC, ORDVAC, ILLIAC Many used vacuum tubes and mechanical relays Personal Computers 1940 Berkeley NOW Project 1950 Server Farms (e.g., Oceano) 1960 P2P Systems (90s-00s) Many Millions of users 1970 Many GB per day 1980 1990 Data Processing Industry - 1968: $70 M. 1978: $3.15 Billion 2000 Timesharing Industry (1975): 2012 Cloud Market Share: Honeywell 34%, IBM 15%, s Grids (1980s-2000s): Xerox 10%, CDC 10%, DEC 10%, UNIVAC 10% GriPhyN (1970s-80s) Open Science Grid and Lambda Rail (2000s) Globus & other standards (1990s-2000s) Cloud computing: Full circle back to time sharing 39  Introduction to Cloud Computing Applications enabled by cloud computing High-growth applications When you startup gains traction, can you keep up? Friendster(2001): Could not keep up with user growth Facebook (2006): $Billion company today Airbnb, Uber, Expedia,... Aperiodic applications How do you deal with sudden load peaks? https://aws.amazon.com/blogs/aws/amazon-prime-day- 2022-aws-for-the-win/ Uber: https://www.youtube.com/watch?v=KxDWs1JRU70 Flipkart website crashed on their “Big Billion Day” sale due to DDOS attack: https://review.firstround.com/navigating-the- leap-from-big-tech-to-startups-advice-from-a-former-google- and-flipkart-exec If you design for peak, how do you deal with low loads? Amazon normal day: 1.3 billion transactions Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 41 /55  Introduction to Cloud Computing Applications enabled by cloud computing(2) On-off applications Scientific simulation using 1000s of computers DNA Nexus and Baylor college of medicine analyzed DNA of more than 14,000 individuals 2.4 million core-hours of computational time, 440 TB of results, 1PB of storage Why not rent computing time to run such one-off experiments? Periodic applications Stock market analysis Mine market data during day Analyze data during night Different computational requirements at different times Dynamic, flexible infrastructure can reduce costs, improve performance Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 42 /55  Cloud infrastructure Types of clouds Public (external) cloud Open market for on demand computing and IT resources Concerns: Limited SLA, reliability, availability, security, and trust Private (internal) cloud For large enterprises with the budget and large-scale IT Hybrid cloud Extend private cloud by connecting it to public cloud Use the local cloud, and when you need more resources, burst into the public cloud Dropbox use case: https://www.wired.com/2016/03/epic- story-dropboxs-exodus-amazon-cloud-empire/ Multi cloud Replicate/partition microservices across public cloud vendors Prevent vendor lock in, resilient to cloud outage Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 43 /55  Cloud infrastructure Cloud adoption All major cloud providers are extending their offering to private and hybrid markets Example: Google Anthos, Microsoft AzureStack Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 44 /55  Cloud infrastructure Know the leaders Raja Appuswamy (Eurecom) Cloud Computing and Distributed Systems 45 /55