IT Challenge: Server Blades

A financial services firm discovers deploying blade technology isn't as simple as it sounds. Here's how it overcame the challenges.

April 29, 2004

7 Min Read
Network Computing logo

Like many financial-services firms located in high-cost city centers, Townsend Analytics suffers from a lack of space. When it comes to managing computing hardware in the firm's high-density data center, Jon Akeson, a procurement evaluator at the company, says Townsend is "squeezing as many pieces of computer equipment as we can in a small area. It's something we try to maximize."

So when hardware manufacturers started announcing blade equipment solutions, designed to improve server density, Akeson says, "We were pretty excited about it, and we looked at several blade products." The "phenomenal advantages" of blade computing would allow Townsend to "double the amount of servers in a given space," Akeson says.

Once Akeson entered the testing phase and closely examined the different product lines, however, he found that deploying blade technology wasn't as simple as it sounded. In fact, it required "costly changes" to the power-distribution infrastructure and mandated that the firm add some networking equipment.

While Townsend bit the bullet and spent the money, Akeson points out that blades have to make your management of servers easier in order to be an advantage. "We feel it may result in a decrease in the cost of ownership in the long run, but initial problems make it impossible to just roll them into production without some extra expenditure," he says.Still, if the blades work as expected, there are exceptional benefits to be reaped. In addition to reduced costs, blade computing provides better performance, more scalability, increased versatility and better systems management, proponents of the technology say. Though the computing architecture is only now starting to take hold, and many Wall Street firms are just beginning to test its parameters, the deployment of blade servers may be about to explode.

According to John Humphreys, a research manager who covers the server space for market research firm IDC in Framingham, Mass., spending on blade servers reached $650 million globally in 2003. He asserts that by 2007, spending on blade technology will be "north of $7 billion." One of the factors driving growth, he says, is that there's now a "nice ecosystem of vendors" - major firms like IBM, Hewlett-Packard and Sun Microsystems have joined the blade battle, along with players like Marlboro, Mass.-based Egenera and RLX Technologies in The Woodlands, Texas.

Servers have evolved from decentralized satellites located throughout an organization to more centralized administration hubs. The traditional method for consolidating servers was to stack them in bulky towers. In recent years, however, thinner servers have transformed those towers into slimmer racks. The setup saves space and provides easier access for administrators, but each server typically requires its own power access, network and switching. And the configuration can be a cabling nightmare. According to industry experts, the use of blade servers, which can eliminate many of those obstacles, is the next stage of server consolidation.

The lingo of blade computing is somewhat ambiguous. A blade server is usually considered the chassis or rack - which is designed to optimize the computing process - that in turn stores a number of server blades, which are hot-swappable devices. Each server blade is essentially an independent server that has one or more processors and its own memory and network controllers. Server blades have their own operating system and run their own applications. Individual server blades slide into the chassis - comprising a blade server - and then share a common infrastructure with other blades. Shared components could include a power supply, fans and cooling systems, Ethernet connections, and switching.

Simply, blade technology allows IT managers to pack more computing power into less space. For example, the Egenera BladeFrame System can hold 96 high-end Intel processors in a 24-inch by 30-inch by 84-inch chassis. IBM's BladeCenter T chassis allows firms to pack 80 processors in an 84-inch rack. Traditional server racks typically hold no more than 42 devices.

Tim Dougherty, IBM's director of blade strategy in Armonk, N.Y., says blades "enable you to simplify your data center." By combining things like servers, switching and Ethernet inside "the same enclosure, you not only get savings in power and space," but it makes it easier to manage the entire system, he asserts. "It's a much easier infrastructure to deploy than the rack-mount servers." And it's cheaper.According to "Why Blade Servers?" by Mark Chapman of the IBM Server Group, blade servers have six major cost-savings advantages over traditional rack servers:

  • Lower purchase price. Blades eliminate the need for duplicate components and can reduce cabling costs.

  • Lower installation costs. Traditional rack servers require assembly, configuration and cabling, much of which is eliminated with blades.

  • Less power consumption. By using low-end processors, blades can save money on power and cooling costs.

  • Space reduction. The increased density of blade computing allows firms to use less space for more equipment.

  • Lower failure costs. Traditional servers require dedicated fans, cables and accessories, increasing the risk of failure. Blades require fewer components, reducing the likelihood of a component failing.

  • Scaled purchasing power. Firms can buy just the computing power they need now and add blades as their processing requirements increase in the future.

    Despite these advantages, Dougherty notes, winning over Wall Street is no easy task. "Not everybody is convinced they see the value of blades. Some people are a little skeptical of taking things and embedding them in the blade center." But once they see the benefits, he says, "They see value."

    John Knuff, senior vice president of network engineering at Nyfix, isn't among the skeptics; he relies on blades to run mission-critical systems. Nyfix, which provides electronic trading and infrastructure technologies to institutional trading firms, processes between 400 million and 1.2 billion shares daily. Nyfix uses blades to power its FIXBox Instant Integrator technology, a hosted plug-and-play FIX solution for buy-side firms that allows them to obtain rapid integration of FIX connectivity into their trading applications. The system uses IBM's BladeCenter eServers running Linux.

    "Blades are best when you've got servers that are dedicated to a specific role and there are six or more," Knuff explains. "If you've got that situation and have expensive data-center space, blades make good sense for faster deployment and economics. If you have unlimited data center space, which most companies don't, it may not make sense for you," he says.

    Townsend's Akeson, who also uses IBM blades, warns, however, that not all blades are created equal. It's important to canvass the options, he says. For example, Akeson had to convert part of his data center to 208-volt power to accommodate Townsend's blade solution. Another vendor's solution would have required Townsend to convert to 208-volt power and add an AC/DC power supply. Akeson says blades are good "when you know exactly what you are going to do with them. If you need to use those resources for different tasks, you may find it's not applicable and you're stuck with the blade servers." Additionally, the blades may only fit a proprietary chassis and may not be compatible with other equipment.Vern Brownell, former CTO at Goldman Sachs and the founder of blade-maker Egenera, which focuses on running mission-critical applications, asserts that the different approaches taken by vendors to build blade technology can cross firms up. Some offerings are simply "miniaturized servers," he says, which compromise the design and are "not really built for core data-center applications." Brownell, whose firm's clients include Credit Suisse First Boston, JPMorgan and Goldman Sachs, warns that a firm could find its network "polluted" by vendor switches that are incompatible with the company's current network.

    Brownell says the financial-services industry is in the early stages of blade adoption and he expects it to grow, especially as Linux takes hold. That will put pressure on firms running UNIX. UNIX boxes are expensive, he says, and firms can "take hundreds of proprietary UNIX boxes and collapse them down into a few of our BladeFrames and get huge cost advantages and better reliability."

    Akeson agrees that IT managers have to shop carefully for blade solutions, remembering that if a firm has to rework its data center to add power, that has to be taken into the total cost of ownership. "The IBM platform works best for us. I wouldn't say that makes it best for anybody else," he explains. It really depends on a firm's needs, he says, adding that, "Each one of the blade packages comes with something unique that could be a huge asset to different companies, depending on the way the operations group runs."

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights