The primary goal of Rocks is to enable non-cluster experts to build and manage their own clusters. To this end, Rocks provides a turn-key distribution that allows one to deploy clusters in an automated fashion. In addition to covering cluster deployment and management using NPACI Rocks, this talk will cover traditional cluster architecture and specify hardware and software components. In this workshop, we present a practical introduction to the Rocks Cluster Toolkit including the design philosophy, front-end/compute-node designations, software “rolls”, and user customization options. In addition to presentations from TACC staff, the workshop concludes with a hands-on lab to build a working cluster. The current release of Rocks 4.0 is based on a CentOS Linux distribution, which is in turn, based on RedHat Enterprise 4.0. This version supports a variety of commodity hardware including x86, Opteron, EM64T, and Itanium processors.
- Background/Philosophy of Rocks
- Master Node Definition
- Compute Node Definition
- Hardware Configuration
- Software Rolls
- Kickstart / XML Configuration
- User Extensions / RPMs
- Hand’s On Lab –Build a Working Cluster