Four steps to buying better blades

The one-box wonder isn’t going to cut it for Nigel Fortlage anymore.

A few years ago, the vice-president of Winnipeg-based customs brokerage GYI International Ltd. became an early adopter of virtualization, consolidating a number of server environments into a single IBM iSeries machine. As GYI approaches its end of lease, however, Fortlage has determined the costs are getting too high, so he’s asking his management to approve the deployment of some blade servers instead. He knows there’s some risk involved.

“Definitely the message from Gartner (at a recent event) was, if you don’t need to do blades today, then don’t,” he said. “They were saying that the investment you make may not have a future and you may be doing forklift upgrades until the standards are there. That put a scare into me.”

The fear isn’t really hurting sales, however. According to a recent IDC Canada market update, spending on the overall server market is fairly weak, with growth forecast at only 1.5 per cent. Within that, volume servers are expected to grow at three per cent. Blade servers, however, are forecast to grow at 10 per cent, according to the research firm.

“There has definitely been some impact on servers due to virtualization and getting more utilization out of infrastructure,” said IDC Canada managing director Vito Mabrucco. “Smaller servers are getting more powerful.”

As a loyal Big Blue customer, Fortlage said he’s sticking with IBM, but that his research had him speaking with several vendors, including Sun and HP, for the first time. “I’ve had some interesting conversations in the last six weeks,” he said.

Although it may take years before you can take any blade and make it work with any other blade or module, there are a few steps you can take now to make sure your deployment goes smoothly. Consider these do’s and don’ts.

More in ComputerWorld Canada

HP takes aim at mid-market data centres

Figure out what kind of blade buyer you are

IT managers should consider ease of growth in their data centre, the upgrade path for the blade servers in question and the number of components they’ll need to deal with, especially for machines with multiple power supplies. Many of the major vendors have created an array of blades for a variety of specific needs.

“If you have a hammer, everything looks like a nail,” says Brian Down, chief technologist and principal engineer at Sun Microsystems. “Well, we have a hammer, a screwdriver and an awl.” In other words, Sun includes a multi-threaded Sparc-based blade, a Xeon-based box and a X64-based machine in its portfolio.

“If you’re a telco, you probably want to stay with one of our Netra (ACTA) product. If you’re running a high-performance computing environment, you should be looking at X64-based products. If you’re looking for throughput, we’ve got the (SunFire) T2000.”

Earlier this year, IBM added to its own lineup with the HS12, a low-cost blade that starts at only $999 and was specifically designed as more of an entry-level product.

“This is specifically aimed at those customers that have standardized on a blade mechanical footprint and want to put all their server platform on a blade,” said Frank Morassutti, business development manager of modular systems at IBM Canada. “They have found their older blades were too big, too expensive, too powerful. They didn’t need that kind of performance for file and print, or for a Web server.”

If customers only have a few servers, however, it may be difficult to justify purchasing a blade infrastructure, said Kevin Smith, enterprise solutions marketing manager at Dell Canada. “We don’t have a ‘blade everything strategy.’ It’s not appropriate for every customer,” he said.

Understand your application mix

Even if you’re using a multi-threaded product, you still have some questions about your software programs to answer, Down says. “Are they floating point- or I/O-intensive? Not everyone understands that about their applications,” he says. “They’re not ready for virtualization so they can’t take advantage of the compute density, or don’t know their apps and what product they should be buying for their box.”

Dell’s solution to the I/O problem is Flex I/0, an upgradeable port that is integrated in its M1000E module for PowerEdge blades. That means customers can add modules like 10 Gigabit Ethernet after the fact. “You don’t have to rip and replace the network technology to get the most out of the investment,” he said. “We’ve created a large amount of I/O bandwidth — 5.2TB per second overall bandwidth. The expansion ports have 80 Gbps per port theoretical capacity. That provides some future-proofing against other technologies like 10 GigE.”

Stay cool

Blade servers can help with data centres in facilities that don’t provide enough power for their existing equipment, and virtualization software can increase utilization, but air conditioning costs are another matter.

“If you make servers more efficient, then it has a cascading effect on the rest of the data centre. You end up with less energy to power the systems or to cool the systems,” Smith says.

Down agrees. “When you go for that kind of density, you’re really putting a lot of kilowatts on a single rack. That creates its own challenges on your data centre infrastructure,” he says. “If you go for high-density products, you want to be concerned about what infrastructure is going to keep those products cool in an appropriate fashion. In some cases, it might make sense to have a single blade in a chassis in a rack.” Single components can help reduce the draw on power, he adds.

There’s also some work being done on the blades themselves. Morassutti points out that IBM last year formed an agreement with SanDisk to put flash drives on blades, which he says has reduced power consumption compared to spinning hard drives.

Get the ROI right

“The focus tends to be on how to make the most out the investment, specifically the investment in the blade chassis itself,” says Smith. “They’re trying to figure out how to maximize that.”

For a blade server with a two or three-year life span, Morassutti says it’s a mistake to calculate that return on investment based on the purchase price alone.

“People that go with blades recognize the value of infrastructure savings — what’s the cost savings on air conditioning, cable manageability, ease of installation? You have to look at the tangible benefits over time.”

Related Download
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center Sponsor: Lenovo
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center
Find out how Hyperconverged systems can help you meet the challenges of the modern IT department. Click here to find out more.
Register Now