- Количество слайдов: 13
Cloud Computing Why is it called the cloud?
• • Clouds and VM First step to learning about cloud is Virtualization Taking VM from your desktop to cloud is our goal Scaling and why it matters to have many VM ? Connecting VM’s and what is an appliance ? Discussion on VM <-> Cloud Why do we need to learn to script and automate our infrastructure ? But what about HPC ?
Cloud Concepts • It is critical to build a scalable architecture in order to take advantage of a scalable infrastructure • Identify the monolithic components and bottlenecks in your architecture • Identify the areas where you cannot leverage the on-demand provisioning capabilities • Refactor your application in order to leverage the scalable aspects • Scalability flavors: • Vertical Scaling • Horizontal Scaling • Elasticity and finding the happy medium • Think parallel and decoupled! • Design for failure. • If its not “automated” you are not doing cloud the rights way !
Cloud: Essential Characteristics • • • On-demand self-service: Client can provision resources as needed in an automatic fashion without human interaction with provider Broad network access: Resources are accessible through the internet Resource pooling: Provider’s resources are pooled to serve multiple clients. Resources can be reassigned as needed Rapid elasticity: Resources can be provisioned rapidly Measured service: Resource usage/allocation is monitored/metered for each client
Cloud: Service Models • • • Cloud Software as a Service (Saa. S): Provides software applications through cloud infrastructure for clients to access through thin-clients (e. g. webbrowser) Cloud Platform as a Service (Paa. S): Provides infrastructure for applications deployed by the client (e. g. provides an operating system) Cloud Infrastructure as a Service (Iaa. S): Provides processing, storage, networks, and other fundamental resources where the consumer is able to deploy and run arbitrary software (e. g. client installs an operating system)
• • Cloud: Deployment Models Private cloud: Cloud infrastructure for exclusive use by a single organization. (e. g. , UA’s cloud for IT services) Community cloud: Cloud infrastructure for exclusive use by a specific community. (e. g. , Cy. Verse cloud, Jet. Stream, Chameleon) Public cloud: Cloud infrastructure open for use by the general public (e. g. , Amazon AWS, Google) Hybrid cloud: Cloud infrastructure that is a composition of two or more distinct cloud infrastructures as listed above that remain unique entities (we will want to be here)
Distributing complete software stacks http: //www. biomedcentral. com/1471 -2105/13/42 8
What is inside ?
What are we using ? • • • Openstack (and the provider is NSF Chameleon) https: //chameleoncloud. org/docs/userguides/openstack-kvm-user-guide/ Quick overview of concepts (review of docs at Chameleon) Review of our “CI Lego blocks” Go to http: //chameleoncloud. org get a account and put it on class roster. I will add you to our class allocations
• • • Ansible Infrastructure as code ! Configuration Management (why ? ) Popular tools for CM (Chef, Puppet etc) Why Ansible ? Concepts: • • Inventory Playbooks (recipe) & YAML Role of ssh and keys ! Ansible: From Beginner to Pro by Heap, Michael (ebook in our library)
Home work: How much does it cost? http: //calculator. s 3. amazonaws. com/calc 5. html
How much does it cost? http: //calculator. s 3. amazonaws. com/calc 5. html • • • You have 2 TB of data. You make the computational data parallel (Different chunks of data may be processed simultaneously). You want to process it, but it takes one core and 4 GB of RAM 10 h to process 200 MB of data (RAM usage scales linearly with data). How much will it cost to process all the data in 1 hour? Note: the final output is 100 MB. How much does cost to transfer the data to AWS and transfer the results back? What is the difference between using the East Coast versus the West Coast facilities?