Cover

Blades demystified

Blade servers promise much, but organizations must be careful to differentiate between tall claims and reality, writes Varun Aggarwal

The growth in the market for blade servers has perhaps been the steepest of any server technology in the recent past. There are numerous benefits that this technology promises to deliver and as many problems that it claims to address. But is it really a technology worth investing in? That’s a question that all blade vendors would like to answer in the affirmative.

As servers proliferated during the past 10 years, the quest for greater server density (that is, more servers per square foot of floor space) to house them emerged. To answer that demand, vendors turned to rack-mountable products. Over time, the vertical space that these servers used was gradually reduced in size to accommodate as many servers as possible in a single rack cabinet. This vertical space was articulated in U height—the amount of vertical distance that a device occupies in a rack cabinet, where 1U equals 1.75 inches. When servers dropped to the 1U height size, it seemed that the minimum size limitation for a single server was reached. To increase server density further, the next logical step was to develop form factors that enabled more than one server to be installed in a single chassis. These servers became known as “blades.”

According to Gartner estimates, blade shipments accounted for 4.3% of overall server shipments in India in 2006. Compare that to the 86% controlled by rack mount servers and you’ll see that blades have a long way to go. Due to growing acceptance of blades, the market share of this technology is expected to reach 11.8% by 2012. That said, rack mount servers would still be way ahead of the blades even five years down the line.

Naveen Mishra, Senior Research Analyst, Gartner said, “We are expecting an environment where we will see all the form factors evolving. Blades have evolved over a period of time and are still evolving.” Blades offer an alternative for space-crunched data centers, better manageability and maintenance.

Ajay Mittal, Country Manager, System x and Blade Centre, IBM India/South Asia explained the reason for the low penetration of blades in the server market, “Blade sales have been focused on enterprises and not small and medium businesses. SMBs would open up a huge opportunity for blades. Now vendors, including IBM, are launching blades for the SMB market space. Consequently, the penetration of blades would witness a steeper growth.”

He added, “Blades house both servers as well as storage. Server consolidation and storage consolidation go hand-in-hand. This was not prominent before and would thus give a further boost to the blade server market.”

The blade chassis

One of the major problems that blade servers are associated with is the vendor lock-in due to the existence of non-standard chassis. “While blade servers are open, the limitation of not being able to use them between chassis from different vendors is based more on the way that they are designed. The blades physically connect to the midplane of the chassis and need an interface for power and I/O connects. Two things that impact this are the physical position of these connects; and the rating of the blades,” explained Anup Gupta-General Manager X64 Systems, Sun Microsystems India.

Mishra added, “Blades have industry standard components, but it’s all housed in a proprietary package in a closed design.” He pointed out that chipsets and backplanes are still evolving for blades.

Gupta agreed with this assertion, “This is true to the extent of the chassis. A customer can always have chassis from different vendors coexisting in his environment. It’s important to avoid an architecture, which is proprietary and ties down your future options. Sun has designed blades with open I/O, unlike comparative blades which have proprietary switching inside.”

HP has an altogether different approach to this issue. Durgadutt Nedungadi, Director, Marketing & Alliances, Technology Solutions Group, Hewlett Packard India, felt that the vendor lock-in is not really a cause for concern for customers. “Though the blade chassis does not allow you to put in a server from another vendor, the so called ‘lock-in’ would only be limited to a maximum of 16 servers as the maximum number of servers that a chassis is designed to house is only 16. Beyond that, the customer is open to buy chassis and servers from any vendor he wants.“

He said that the customers today are more concerned about power and cooling management of a data center rather than the chassis, which does not cost so much.

Power and cooling

Power and cooling of data centers have become a prime focus area for IT managers. Inefficient management of power and cooling may lead to a loss of thousands of dollars.

Blades require more power per square feet of area in a data center, pushing further the thermal envelope of the data center. The extra heat generated demands further cooling. Before deploying blades in an existing data center, it is important to assess whether the thermal envelope of the data center is in a position to take the extra burden.

Nedungadi conceded, ”It’s true that the power consumption of blades is more than that of rack mount servers in terms of per square feet area, but a blade chassis would consume much less energy than a rack mount filled with the same number of servers.”

He said that thermal assessment of data centers should be done in any case in order to identify the hot and cold aisles. This helps in diverting the airflow in such a way that the data centers are able to run with minimum cooling requirements, reducing the power bills to a great extent. Once the thermal assessment is done, it becomes much easier to deploy blades in your existing data center.

The right fit?

Often this question arises if blades fit into all situations. According to Mishra, blades are not the answer for everything. Users should not get carried away by marketing pitches. “Users should look at a shorter ROI perspective of about two-three years as the technology is rapidly changing. If you’re investing in a longer ROI, then the components may not even be available at a later stage.”

Mishra pointed out that users needed to understand the compatibility of blades with their storage and network infrastructure. “They should not assume that a blade would work in their existing infrastructure.” He believed that users should assume the same throughput with blades as they do with rack mount servers since the latter are already mature while blades are still evolving.

Nedungadi felt that there were only two situations in which blades did not fit in. “Blades don’t fit into an environment where you have only two-three servers and you know that your requirement will not increase in a few years. In such a case investing in blades costs you more.” The other situation where one would want to invest into rack mount servers instead of blades would be for example, when you are looking for specific ports or a specific PCI interface.

Though blade servers are a promising technology, it may not satisfy your expectations if it is deployed without due diligence. It is imperative to justify your investment in blades through proper assessment of the technology for your data center.

varun.aggarwal@expressindia.com