The open source IT research and advisory community Wikibon.org released results from its latest Peer Incite research advising users on blade server investments.
According to research from Wikibon.org analysts, blade servers don't perform well in large scale transaction processing situations, vendor lock-in is an issue with the chassis systems, and for applications that require less than 5-10 servers, blades aren't economical.
The open source community analysts - which includes researchers, technologists, and consultants from respected organizations such as IDC - caution users not to rush into blade server computing, despite mantras like "blade everything" from Hewlett-Packard Co. (HP), manufacturer of the HP BladeSystem, because the benefits come at a premium and could double or triple the cost of a server complex in certain cases.
"I would advise users to know their applications and recognize that suppliers want you to buy technologies on their schedule, which doesn't necessarily coincide with yours," said Peter Burris, analyst and chief content officer at Wikibon. "Servers are reliable and keep cranking along, so vendors look for ways to encourage customers to turn over hardware."
With applications, such as email and Web serving, blade computing may allow for simpler configuration, change and operations of hardware, Burris said. But configuring very large transaction processing applications on specially configured hardware is just as challenging with blade platforms as it is with other platforms, he said.
"Blade computing promises many benefits, but eradicating the requirement for labor-intensive software configuration in complex application environments is not one of them -- despite supplier claims to the contrary," Burris said. "This is especially true in server consolidation projects. While blade computing can mitigate many hardware and hardware administration costs, the main challenge of any program to reduce server counts remains application and not hardware, consolidation."
Joe Clabby, head of Clabby Analytics, said Burris' estimation is on target. "High read/write ratios may not be best suited for blades or racks because of bus speeds, memory limitations, disk access, network I/O and any of a number of other factors," Clabby said. "It most assuredly is difficult to build a large scale transaction processing environment made up of tens, hundreds, or thousands of blades front ended by other processors to delegate workload and monitor transactions."
HP, for one, contests the view that blades aren't well suited for large-scale computing. Steve Gillaspy, group manager, HP BladeSystem, said the HP BladeSystem c-Class server blades address any workload -- from simple Web serving to complex database transactions -- with no compromises in capability or performance. "We offer a wide variety of two and four socket blades … to address the performance characteristics of varying application workloads," Gillaspy sajd.
HP server blades are designed for applications including Web serving; infrastructure applications such as DNS; Active Directory; file serving; and terminal services, Microsoft Exchange and Lotus Notes, small and large database applications such as Microsoft SQL Server and Oracle; enterprise resource planning (ERP) applications such as SAP; and virtualization applications such as VMWare and Microsoft Virtual Server, Gillaspy said.
Blade server lock-in
Despite some debate over their value and utilization, users are buying blades, and IBM reports blade servers are now outselling its 1U rack servers.
Over the past few years, large server vendors, like Dell Inc., HP and IBM, have focused on blade technologies; the packaging of CPU, memory, network and I/O capabilities, utilizing common sets of technologies that can be easily added to or removed from a shared frame and controlled with a single set of management resources.
While these vendors use similar nomenclature and concepts, different vendors' offerings are still distinct enough that they can't be mixed and matched. Blades from different vendors have different "ecosystems," and blades from one vendor will not fit into another vendor's chassis.
This is a fact that won't likely change anytime soon, according to IBM's world wide marketing manager of IBM's BladeCenter Scott Tease. "There has been a push toward standardizing blade chassis, but I don't ever see that happening," Tease said. "I couldn't imagine IBM getting together with Dell and HP to share chassis. It is what makes each unique."
HP blades are designed to only run in the company's own chassis, and the design and virtual connect innovation in the backplane of the c-Class chassis is a key differentiator for HP, said Gillaspy.
Blade tipping point
With that in mind, Wikibon warns that small businesses should carefully assess the marginal benefits versus the costs of using blade server technology. Typically, application loads requiring less than 5-10 blades (10-20 CPUs) are best devoted to standalone server technologies, the the open source community's report states.
Users pay a premium for lower volume blade complexes to cover the cost of blade chassis and packaging technologies, as well as vendor markups for their specific solutions, Burris said. He also said users will benefit from blades if they plan on filling a blade chassis and have limited data center space to work with.
"This is excellent advice," Clabby said. "If you don't fill the blade enclosure, you're not maximizing the payback." The IBM blade chassis, for example, holds 14 blades and costs $2,800. When comparing 1U servers and blades, it becomes more cost effective to go with blades if there are more than eight servers, IBM's Tease said.
"Blades are basically a packaging advance, but they are still dependent on the physics of processing," Burris said. "Before going with blades, know the scale of your workload today and for the future."
Let us know what you think about the story; email: Bridget Botelho, News Writer
Also, check out our news blog at serverspecs.blogs.techtarget.com