Software

Penguin Looks To Conquer HPC Clustering

Enrico Pesatori knows a thing or two about business strategy. He has been admired for his self-sustaining executive track record at Digital Equipment, Compaq and BlueArc. Now, as chairman and CEO of Penguin Computing, he’s out to make the company a leader in high-performance computing servers, systems and cluster-management software.

When the company was founded in 1998 by Sam Ockman, a Linux pioneer who was part of a group that coined the term “open source” and is now Penguin’s Director, it was primarily known as the Linux user’s source for server hardware. Today, Penguin’s offerings range from utility servers to enterprise servers to clustering-optimized servers. Unlike other Linux box vendors, Penguin not only survived the technology fallout of the 1990s but became scrappier than ever in spreading its Linux might around, from Fortune 500 customers to its newest target, departmental users seeking HPC clustering.

“The biggest milestone since our founding is our focus on clustering,” Penguin vice president Mark Walker told LinuxInsider. “We will make a name for ourselves on clustering. We still have a significant presence in the enterprise with a lot of the Fortune 500 companies as customers, but you will see us grow, particularly in the area of high-performance clustering.”

Clustering in a Box

Penguin’s BladeRunner “cluster-in-a-box” server, unveiled in December, is Penguin’s way of cornering a lucrative market opportunity in the lower end of the marketplace where HPC appetites are clouded by a distaste for time-intensive configuration and expensive administration. Using very thin blade servers in a clustering design has become a popular method for saving computing resources.

The term “cluster-in-a-box” is intended to be taken literally. The BladeRunner server is essentially hardware with built-in clustering software written by Scyld Software, founded by Don Becker and acquired by Penguin Computing in 2003. Becker, a former NASA scientist and a pioneer in Beowulf clustering, is now Penguin’s CTO.

Scyld develops what Aberdeen Group analyst Bill Claybrook calls the “best-of-breed” Beowulf Linux cluster management operating system. The BladeRunner server is pre-installed with Scyld Beowulf, which means users get a single point of installation, login, and administration. It is a tool for painless integration of the entire blade operation — including servers, Ethernet switches, storage subsystems, management software, and cluster operating system software in a single 4U chassis.

Penguin’s Web site also points out that the BladeRunner cluster-in-a-box is “easily scalable” by adding additional 4U chassis and connecting the integrated Ethernet switches. Its components offer redundancy and hot-swap capability. Prices range from US$23,400 (five-node) to $39,800 (11-node).

Market Explosion

Penguin’s Walker told LinuxInsider that the blade business is about a $1.2 billion market this year, and IDC predicts it will be an $8 billion market by 2008. “It is a very fast-growing market, and Penguin is jumping in with both feet,” he said.

The blade approach — which uses super-thin servers that take up less space — is heavily targeted at data-center managers. The sales pitch for the blade approach is that it consolidates computing power and reduces system-management costs. Sizing up or down is easy, too, which means managers can adapt computing needs as business peaks come and go.

Clustering comes in after packing the extra-slim servers into the data center; the servers are “clustered” with software to create a system of pooled computing resources.

This is where BladeRunner’s edge makes a difference: Traditionally customers have complained that clustering is difficult to accomplish because configuration is challenging. BladeRunner is selling on the strength of its user-friendly characteristics. “Our primary message is ease of use,” said Walker. “One can monitor the different components: the power supply, the temperature of the fan, the fan speed. BladeRunner allows you to manage the nodes in a better way.”

Fast-Growing Segment

The cluster-in-a-box may be appealing to blade enthusiasts at data centers or to corporations with large server farms, but Penguin says one of its most important targets is in “departmental clustering.”

“This is not the million-dollar cluster that a company is deploying but the departmental cluster,” Walker said. He referred to IDC analyst Chris Willard’s breakdown of the clustering market into four components.

“Basically, if you were to break it down in price, it would be zero to $250,000, which is departmental clustering; and then $250,000-$500,000; $1 million; and above $1 million,” said Walker. “In this departmental space lies 46 percent of the market. It’s the fastest-growing [segment].”

Competitive Talk

Walker says BladeRunner outdistances competing products in a number of ways: compute density, power density, manageability and competitive pricing. Its Intel Xeon CPUs offer environmentally friendly features such as low voltage, which means less cost is incurred when having to cool large server farms.

Anatomy of a BladeRunner:

  • 4U (7″) Rackmount Chassis
  • Up to 12 Individual Blades
  • Up to 24 Intel Xeon LV Processors
  • Up to 48GB of PC2100 ECC Reg. DDR RAM
  • Up to 4 Terabytes of Storage
  • Up to 2 Switches w/ 16 External Gigabit Ethernet Ports
  • Up to Two 24-Port Ethernet Switches
  • PCI-X Mezzanine Card (PMC) Expansion Interface
  • Redundant 3+1 Power Supplies

Leave a Comment

Please sign in to post or reply to a comment. New users create a free account.

LinuxInsider Channels