Server Makers Push Linux As Linux Pulls Them

by Timothy Prickett Morgan

The bi-annual LinuxWorld trade show is under way this week in San Francisco, and once again the major platform makers are swearing their fealty to Linux and trying to figure out any angle they can to use Linux as a lever to wrest sales away from their competition and money out of their customer bases. The Linux market has largely proven itself (even if it is still dwarfed by Windows and Unix), and now the major server makers are getting down to the brass tacks of marketing as well as broadening and deepening their support of Linux on their platforms.

Over at IBM, which could turn out to be the vendor that is driving more Linux-related sales than any other player in the IT market, the company's major announcement at LinuxWorld is a new offering that it calls the "Grid to Grow" package. According to Al Bunschaft, vice president of grid and virtualization technologies at IBM, the idea is to break through the paranoia barrier that many customers have in adopting grid technology. Oddly enough, even though about two-thirds of IBM's grid-based sales engagements have Linux as a component (that is spanning all of its eServer platforms, not just X64-based xSeries servers), Linux is not really a barrier to the adoption to grid. The perceived--and misconceived--disruption of adding grid technology is the major barrier. So, hence, the Grid to Grow offering is a low-cost bundle of software that allows companies to test out grid technology and then grow from there.

"We're seeing two things out there in the market," explains Bunschaft. "Too many customers are concerned about the complexity of introducing grid technology in their environments, and at the same time, we know from our customer engagements that grid is a great way to add capacity to existing environments." Grid technology is not about rip and replace, but add and assimilate.

The Grid to Grow offering is not available on rack-mounted servers or tower servers, but only on IBM's BladeCenter blade servers. For a mere $49,000, IBM takes a BladeCenter chassis that can house 14 two-way blade servers and fills it half-way up with seven blades. Based on past sales, this is, says Bunschaft, a typical entry configuration for a company that is testing out the idea of grid computing. To this, IBM adds Red Hat Enterprise Linux 4 on each blade, and then throws in the customer's choice of a workload manager for the grid software: Altair's PBS Professional (a commercial version of the OpenPBS, or Portable Batch System, grid job scheduler), DataSynapse 's GridServer, Platform Computing's LSF, or IBM's LoadLeveler. Bunschaft says that the variety in schedulers is necessary because different schedulers are used for different grid applications or in different industries.

The default configuration in the Grid to Grow offering has RHEL4 and Altair's PBS for that $49,000, and has seven Xeon-based HS20 blades, each with two processors. Customers can opt for the Opteron-based LS20 blades or PowerPC-based JS20 blades, and they can also pick Novell's SUSE Linux Enterprise Linux Server, Microsoft's Windows, or IBM's AIX for the blades (provided the blades support those operating systems; Windows is only available on Xeon and Opteron blades, AIX only on Power blades, and Linux spans all blades). Pricing will be different for these configurations. On top of this hardware, operating system, and middleware stack, IBM throws in project planning, implementation, and deployment services from its Integrated Technology Services unit of the Global Services group. In fact, the whole shebang is sold through the Global Services organization. The offering will be available through IBM Global Services in the Americas region immediately, and will be launched in the rest of the world starting at the end of the third quarter.

In a separate announcement, IBM will also talk about the 4.5 teraflops BladeCenter Linux-Power cluster that it has sold to New York University, which will be the largest supercomputer on the island of Manhattan. This cluster is comprised of 256 of the JS20 blades using 2.2 GHz PowerPC 970 processors, and will be notable in that it will be the first commercial supercomputer that makes use of the IPv6 protocol. This supercomputer, which is being financed in part by the U.S. Army, will be used for genomics and atmospheric research. The deal also calls for a 64-node JS20 cluster rated at about 1 teraflops to be installed as a test and development system. NYU picked the BladeCenter for its cluster partly because of performance, but also because of the density and the high-cost of New York real estate. The 4.5 teraflops cluster fits in a 30 square feet footprint.

Over at Hewlett-Packard, this LinuxWorld is all about point solutions that flesh out its Linux ecosystem. HP launched its "BigTux" extensions to Linux that allowed it to scale across as many as 64 processors in an Itanium-based Integrity Superdome server. Today, HP will announce that it has a customer--Yoon's English Academy in Korea--that is replacing mainframes with such a big Linux box.

HP will also announce that the Virus Throttle technology that its HP Labs division created to choke back viruses on systems to minimize the damage they can do has been ported from Windows to Linux. This Virus Throttle software will be delivered in September for Linux through the ProLiant Essentials Intelligent Network Pack - Linux Edition. At the same time as this is delivered, HP will ship a new version of the ProLiant Essentials Rapid Deployment Pack - Linux Edition 1.3, which supports new ProLiant hardware and new versions of Linux from Red Hat and Novell.

Jeffrey Wade, worldwide open source and Linux marketing manager at HP, says that HP will launch four new Linux Expertise Centers to help independent software vendors in the United States and Europe port their applications to Linux from other platforms. These centers are packed with the latest HP iron as well as Red Hat and Novell Linuxes. ISVs who don't want to travel to these centers to do their porting can use virtual server capacity offered remotely, or they can acquire systems of their own from HP at discounts given only to application partners. HP will also set up a network of over 80 HP Solution Centers to help customers and their partners port their applications to Linux and test them before they roll into production; these centers will be located in major metro areas around the globe.

Wade says HP will create a Linux Reference Architecture--shorthand for a bill of materials to create a complete system including servers, operating systems, and middleware all certified to interoperate--for the OpenLDAP open source directory server. HP will also announce it has signed an agreement with Symas, the main contributor to OpenLDAP, to certify OpenLDAP and the commercial version Symas sells, which is called Connexitor Directory Services, on HP's Linux boxes and LRA configurations. HP will also apparently announce that certain LRAs will now be supported on HP-UX servers based on the Itanium platforms and running within the Linux Runtime Environment (LRE) that has been part of HP-UX on Itanium from the beginning. LRAs based on JBoss, MySQL, and OpenLDAP will be the first three LRAs to be ported to the LRE environment within HP-UX.

And finally, HP will announce it has over 200 open source programs that are now supported on its Unix-oid NonStop fault tolerant server platform, which was recently moved from the Silicon Graphics MIPS processor to an Itanium architecture that is common to the Integrity servers that support Linux, HP-UX, Windows, and OpenVMS. These programs include the Apache Web server, the Samba print and file server, the Zope content management system, and the Jabber blogging system. Wade says that the company has another 300 open source programs being ported to the NonStop platform in the coming months.