A General Purpose High Performance Linux Installation Infrastructure
With more and more and larger and larger Linux clusters, the question arises how to install them. This paper addresses this question by proposing a solution using only standard software components. This installation infrastructure scales well for a large number of nodes. It is also usable for installing desktop machines or diskless Linux clients, thus, is not designed for cluster installations in particular but is, nevertheless, highly performant. The infrastructure proposed uses PXE as the network boot component on the nodes. It uses DHCP and TFTP servers to get IP addresses and a bootloader to all nodes. It then uses kickstart to install Red Hat Linux over NFS. We have implemented this installation infrastructure at SLAC with our given server hardware and installed a 256 node cluster in 30 minutes. This paper presents the measurements from this installation and discusses the bottlenecks in our installation.
- Research Organization:
- SLAC National Accelerator Lab., Menlo Park, CA (United States)
- Sponsoring Organization:
- USDOE Office of Energy Research (ER) (US)
- DOE Contract Number:
- AC03-76SF00515
- OSTI ID:
- 799076
- Report Number(s):
- SLAC-PUB-9193; TRN: US0204385
- Resource Relation:
- Other Information: PBD: 17 Jun 2002
- Country of Publication:
- United States
- Language:
- English
Similar Records
High Performance Diskless Linux Workstations in AX-Division
Petascale Data Storage Institute (Final Report)