Open Stack

Summary

The lecture introduces OpenStack, an open-source cloud platform used primarily to provide Infrastructure as a Service (IaaS). It pools compute, storage, and networking resources in a data center. Administrators manage the infrastructure, while users can provision virtual machines and other resources through a web dashboard or APIs.

OpenStack can be used for both large deployments and lab-scale experimentation. It can be installed on servers or a small number of PCs. The lecture cites Meghamala, IIT Kharagpur’s experimental OpenStack-based cloud running on blade servers, as well as smaller installations used for student projects. These environments can support research into cloud resource management, power management, and “green cloud” systems.

Main OpenStack Services

• Nova — Compute: Manages the lifecycle of virtual-machine instances, including scheduling, launching, and decommissioning them. Users can request VMs with different configurations, or “flavors.”
• Neutron — Networking: Provides networking services and APIs for defining networks and connecting them to instances. It supports internal cloud networking and connectivity to external networks, with pluggable support for different technologies and vendors.
• Cinder — Block storage: Provides persistent block devices that can be attached to running VMs. The storage remains available until a user deletes it.
• Swift — Object storage: Stores and retrieves unstructured data through a REST/HTTP-based API. It uses replication and a scale-out architecture for fault tolerance rather than functioning like a conventional file server.
• Glance — Image service: Stores and retrieves VM disk images, which Nova uses when creating instances.
• Keystone — Identity service: Handles authentication and authorization, and provides a catalog of endpoints for OpenStack services.
• Horizon — Dashboard: Provides a web-based self-service interface for tasks such as launching instances, assigning IP addresses, loading images, managing users, and configuring access controls.
• Ceilometer — Telemetry: Collects measurements for billing, benchmarking, scalability analysis, and statistical monitoring. These measurements can also help operators understand usage patterns and plan infrastructure capacity.

Storage Types

The lecture distinguishes three storage models:

• Ephemeral storage: VM-local storage that lasts only as long as the instance. It is used for the operating system and temporary or scratch space.
• Block storage: Persistent storage attached to a VM and managed by Cinder.
• Object storage: Files and other objects stored independently of a VM and accessible from elsewhere, managed by Swift.

Request and Provisioning Flow

A typical VM-creation sequence begins when a user signs in to Horizon and submits VM parameters. Keystone authenticates the request and returns a token. Horizon then sends the request to Nova. Nova’s scheduler selects a suitable compute host, while Neutron configures networking, Cinder can provide block storage, and Glance supplies the selected image. A hypervisor ultimately runs the instance.

The instructor also outlines internal provisioning mechanisms, including Nova’s scheduler, message queues, database lookups, and communication between Nova components. This is presented as an architectural overview rather than a detailed implementation guide.

Demonstration Mentioned

The lecture promises a live demonstration using IIT Kharagpur’s OpenStack installation, including how to create VMs and use the dashboard. However, the supplied subtitles end just as the instructor introduces the demonstration, so they do not include the walkthrough.

Speakers and Sources

• Speaker: An unnamed lecturer presenting a cloud-computing course.
• Sources mentioned: OpenStack’s website and other OpenStack resources; IIT Kharagpur’s experimental OpenStack cloud, Meghamala.