March 6, 2000
|
Printer ready |
By Jason Levitt
Linux clusters are available in two basic forms. The most mature Linux clustering technology comes from the Beowulf Project (www.beowulf.org), a massive parallel-processing solution geared toward the scientific community, which often needs supercomputing-class performance for number crunching. For businesses that need high availability and load balancing for their Web, FTP, E-mail, and other services, software from the Linux Virtual Server Project (www.linuxvirtualserver.org) and the High Availability Linux Project (www.linux-ha.org) is available in some Linux distributions.
Red Hat's Piranha add-on software (www.redhat.com) and TurboLinux's TurboCluster Server (www.turbolinux.com) are based on the Linux Virtual Server Project kernel patches that makes it pretty easy to put together a two-node load-balancing front end, but it's missing features that some users might find essential. Linux Virtual Server clusters serve up static Web pages and files via FTP, but they don't handle managing dynamic content. They also can't retrieve existing user connection and session information (failover) when a machine in the cluster goes down.
A more mature, but similarly limited, alternative is offered by Penguin Computing (www.penguincomputing.com) for its Intelligent Node clusters. Penguin uses Resonate's Central Dispatch high-availability clustering software for Linux and resells it with its Intelligent Node Linux boxes. Central Dispatch offers superior load-balancing management, but still doesn't have failover capability. It's a commercially proven product, yet adds about $2,000 per machine to the price of the Penguin Linux boxes.
For scientific users, as well as number crunchers in banking and other industries, software from the Beowulf Project lets you create a high-performance massively parallel computer out of commodity hardware PCs and freely redistributable source code operating systems such as Linux and FreeBSD. The problem, of course, is that in order to take advantage of Beowulf's parallel processing architecture, your applications have to be written--or in the case of existing business applications, rewritten--to take advantage of parallel processing techniques, and few business applications are. Still, various commercial distributions of Beowulf software exist, such as VA Linux's ClusterCity system (www.valinux.com) which includes custom, preconfigured, rack-mounted Linux systems and open-source cluster-management software.
he Linux community doesn't have high-availability clustering as mature as NetWare's, but for those who need supercomputer speed and number crunching, or who would like an inexpensive server farm for their Web presence, there are some Linux solutions. Depending on your application, a cluster of Linux systems running on commodity PC hardware and inexpensive open-source software can get you some clustering capabilities at a fraction of the cost of many commercial systems. Still, clustering is immature on Linux systems compared with NetWare, Windows NT, and mainstream Unix systems such as Sun's Solaris or Compaq's Tru64. As on other fronts, though, the Linux community is moving forward with aplomb, and some Linux clustering offerings have arrived. What's required, as usual, is the willingness to put the pieces together yourself.
Return to main story, "NetWare Cluster Services Expand On Reliability."
Back to This Week's Issue
Send Us Your Feedback
Top of the Page
This Week's Issue
Technology Whitepapers
- Mobile BI: Actionable Intelligence for the Agile Enterprise
- Creating the Enterprise-Class Tablet Environment - by Yankee Group
- How To Regain IT Control In An Increasingly Mobile World - by BlackBerry
- Red Alert: Why Tablet Security Matters - by BlackBerry
- New Visual and Wizard-Driven Paradigms for Exploring Data and Developing Analytic Workflows











