November 11, 2011 | Written by: Joydipto Banergee
Share this post:
Update: Read part two and part three of this series.
Introducing the design and implementation of an on-premise IaaS/PaaS cloud
Summary: In this three-part blog series, the author outlines the process, from conception to deployment, that his team used to build a private, on-premise cloud environment by using an IBM hardware and software stack.
The following topics are covered in this part:
- The five phases in the roadmap: from concept to deployment
- Several details of the unique solutions designed for this project
- The typical cloud structure
- The software and hardware requirements list for this project
The five phases of this solution roadmap
The project implementation is in five phases. The next figure captures the overall journey from the initial thought process to the final deployment, divided into the five phases.
Foreshadowing: Several solution details
The team first developed a service view to conceptualize the separation of the consumer and the service. The next figure shows the cloud service view.
Cloud end users should be able to put a service request through a user interface (UI); for example, a WebSphere Application Server (indicated as WAS in the figure) on the AIX OS or WebSphere Portal on AIX with specified computing power, memory, and storage
The cloud operations team would manage the cloud management platform, support infrastructure and operation such as defining the service, publishing the service, analyzing the reports, generating bills based on the usage, capacity planning, and more.
The team decided to use IBM Tivoli Service Automation Manager product suite, considering such other alternatives as Citrix and Open Source Cloud software.
From studying the various pros and cons in hypervisor selection, the team chose IBM PowerVM for cloud image provisioning to use the existing Power hardware infrastructure and to better match the requirements of Tivoli Service Automation Manager.
The typical private cloud structure
Before diving further into the implementation, let’s look at a typical cloud structure for reference. The next figure shows the typical cloud architecture.
In its simplest form, Tivoli Service Automation Manager (indicated as TSAM in the figure) topology consists of one management system server (can be System p, System x, or System z), one administrative server (System x), and one managed server (can be System p, System x, or System z). The management and administrative servers are required exclusively by Tivoli Service Automation Manager for the installation of cloud software; the managed environment involves provisioning and management of the virtual servers based on customer requests.
In this example, one System p server is being virtualized into multiple logical partitions (LPARs) with one LPAR being used for the Tivoli Service Automation Manager management server and possessing Tivoli Provisioning Manager and various middleware products such as DB2, WebSphere Application Server, HTTP Server, LDAP (these middleware products are actually part of the Tivoli Provisioning Manager suite).
The other LPARs of this server are designated for associated but optional components such as IBM Tivoli Usage and Accounting Manager for metering and IBM Tivoli Monitoring for infrastructure monitoring. Two other LPARs are meant for Network Installation Manager (NIM) server (image storage) and VIOS for AIX partitioning.
The other System p server is used for cloud-managed environment where all virtualized images or resources would be provisioned automatically by the service requester, the consumer. (You may use System x for similar provisioning of Windows and Linux resources with VMware hypervisor.)
Another System x server is used for a Tivoli Service Automation Manager administration component with Tivoli Provisioning Manager web, image library, and for Service Request Manager (SRM).
System p hardware is being managed by Hardware Management Console (HMC) hardware as a standard Power system management tool.
All the System p LPARs ran AIX 6.1; the System x ran SUSE Linux 10.2 (both 64 bits). Tivoli Storage Manager was employed to make a backup of the AIX environments; G4L was used for the Linux.
The major hardware and software players for the project
The following list shows the final hardware and software requirements:
- IBM System p 570 for Cloud Management Environment
- IBM System p 570 for Cloud Managed Environment ( for users )
- IBM System x 3850 for Cloud Administration Environment
- Tivoli Service Automation Manager
- Tivoli Provisioning Manager
- Tivoli Service Request Manager
- Tivoli Monitoring
- Tivoli Usage and Accounting Manager
- Hypervisor – Power VM
- For taking image backups: Tivoli Storage Manager, G4L (Open Source)
Note that Tivoli Monitoring and Tivoli Usage and Accounting Manager are optional software components; they are not part of the standard Tivoli Service Automation Manager product suite.
In this article, I’ve provided the background planning concepts for a real-world project implementation to build an on-premise IaaS/PaaS cloud.
Part 2 of this series will cover installing and configuring the components, and several special features of the components.