In Today's post, we continue our discussion on OpenStack. In particular, we refer the Architecture for OpenStack on Ubuntu. OpenStack is described on http://docs.openstack.org It is designed to be massively scalable horizontally. To begin with we can have a single cloud controller that hosts the databases, message queue service, authentication and authorization service, image management service, user dashboard, and externally accessible API endpoints for OpenStack services. By starting with a single cloud controller, we favor simplicity over fault tolerance and availability that is possible with a fully redundant cloud controller. Most OpenStack Compute central services communicate with each other using the Message Queue which also has cluster capabilities. Databases save stateful information and they are critical to the central services. The content and delivery services consists of two parts - one that is responsible for the delivery of images and the latter maintains the metadata information associated with the virtual machine images and requires a database.
The OpenStack dashboard is implemented as a Python web application that runs in Apache httpd. The OpenStack Identity service allows identity to be set in the policy.json file. The identity service supports different plugins for storing information which include an in-memory Key-Value store, SQL database, PAM and LDAP. OpenStack implementations usually involve MAAS and Juju. MAAS is a tool that helps manage physical infrastructure with the same ease and flexibility as virtual machines in the cloud. Specially, MAAS allows us to discover, commission and deploy physical servers and dynamically allocate physical resources to match workload requirements and retire servers. Juju is a service orchestration tool which allows the administrator to configure, manage, maintain, deploy and scale cloud services (workloads) quickly and efficiently leveraging MAAS. Generally MAAS and Juju GUI run on separate servers. Similarly the Controller node, Compute services, Object Storage services and Block Storage operate on separate nodes. This is done to provide dedicated hardware for each of these services and to reduce contention. The Controller and compute nodes should have 4 1TB physical drives with RAID5 striping. The storage should have 2 500 GB physical disks locally in addition to access to an array of such disks.
The OpenStack dashboard is implemented as a Python web application that runs in Apache httpd. The OpenStack Identity service allows identity to be set in the policy.json file. The identity service supports different plugins for storing information which include an in-memory Key-Value store, SQL database, PAM and LDAP. OpenStack implementations usually involve MAAS and Juju. MAAS is a tool that helps manage physical infrastructure with the same ease and flexibility as virtual machines in the cloud. Specially, MAAS allows us to discover, commission and deploy physical servers and dynamically allocate physical resources to match workload requirements and retire servers. Juju is a service orchestration tool which allows the administrator to configure, manage, maintain, deploy and scale cloud services (workloads) quickly and efficiently leveraging MAAS. Generally MAAS and Juju GUI run on separate servers. Similarly the Controller node, Compute services, Object Storage services and Block Storage operate on separate nodes. This is done to provide dedicated hardware for each of these services and to reduce contention. The Controller and compute nodes should have 4 1TB physical drives with RAID5 striping. The storage should have 2 500 GB physical disks locally in addition to access to an array of such disks.
No comments:
Post a Comment