Data centres are bulging at the seams. As organisations have grown, they have added more servers to deal with the extra workload. The net result is that data centres have become difficult to manage and costly to maintain. Vendors believe that a new, smaller form of server, known as the blade server, can solve this increasingly common problem.
In essence, a blade server is a server contained on a card. They are much smaller than standard ‘rack’ servers and can be packed into a far denser area. Some vendors claim they can squeeze as many as 20 blades into one ‘enclosure’, or rack shelf. This has led some analysts to term the blade-server phenomenon ‘ultra-dense computing’.
Unlike rack servers, most blades are comprehensive computing systems. They are typically based on either the Windows or Linux operating system, and the processor, memory, network connections and associated electronics are contained on a single motherboard. Networking, storage and systems management ‘blades’ sit in an enclosure alongside blades serving applications or web pages.
The main selling point of blade servers is ease of management. Using sophisticated management software, systems administrators can install and configure specific blades to serve particular applications remotely and in a matter of minutes.
Established hardware vendors and innovative start-ups populate this embryonic market. Of the major server manufacturers, only Hewlett-Packard (HP) and Compaq Computer have brought blade products to market to date. But IBM, Dell Computer and Sun Microsystems are expected to follow during the second half of 2002.
- Hewlett-Packard claims to have been developing blade servers before the term was coined. It released its first product in December 2001.
- Compaq released its ‘Quickblade’ product – the first in a series of blade server products – in January 2002. The company says 20 of these single-processor blades can sit side-by-side in a single enclosure, which are typically about 60cm wide.
- Dell prefers ‘modular computing’ to the blade tag. It plans to launch its first such product, beginning with both two- and four-processor machines, in the second half of 2002.
- IBM is opening up a two-pronged sales strategy. At present, it resells ultra-dense, low-cost servers from US start-up RLX Technologies. But it is also developing its own blade server product Xcaliber, and these are due for launch in mid-2002.
- Sun Microsystems plans to ship its first blade server product, based on its Solaris operating system and Sparc processor, at the end of 2002. Sun is also developing software to manage these devices, which it will roll out across these products over time.
The new players
A number of start-ups with different approaches to the blade server proposition have attracted significant amounts of venture capital and have poached senior management from established vendors. Although they are unlikely to displace the likes of HP and IBM, their innovative technology could ultimately make them attractive acquisition targets.
- Flagship product: ManageSite blade server platform
This is a combination of ultra-dense hardware and the company’s ‘Virgo’ server reference design, which manages and allocates server resources across dispersed hardware infrastructures.
Funding: Not available
- Flagship product: BladeFrame
Unlike most of its rivals, Egenera’s is targeting the high end of the server infrastructure market. It claims to offer up to 10 times the performance of a high-end Unix server at a quarter of the overall purchase and running costs.
Funding: €57 million
- Flagship product: OpForce Infrastructure automation suite
Jareva focuses purely on sophisticated systems management software for managing large numbers of dispersed servers. Its CEO and CFO are both former senior executives of HP.
Funding: €12.5 million
- RealScale Technologies
- Flagship product: iCluster blade servers and RealScale Manager software
The company began life as a France-based research project into server proliferation. It believes that a blade is just another component that can be managed as easily as a network switch or router.
Funding: €6.8 million
- RLX Technologies
- Flagship product: ServerBlade 633/324
One of the early entrants into the blade server market, RLX’s focus is on ‘redefining server economics’. Despite its substantial funding, the company had a difficult 2001, hit by a spending slowdown and managerial departures.
Funding: €67 million
Despite their apparent benefits, blade servers are still very much in the early adopter phase. Worldwide shipments of blade servers will amount to 900,000 units in 2002, according to Gartner, the market research company, and will not surpass the one-million-a-year mark until 2006. Organisations are holding back on implementing blade servers due to worries about interoperability between servers and management software from different vendors, according to Gartner analyst Adrian O’Connell. They are also confused about how the new technology will fit into their existing technology infrastructures, he says. “The demand for front-end web servers has been critical in driving the need [for blade servers], but in reality the market is only just beginning to understand how they might be deployed and used.”
What are the benefits?
Space: By reducing the amount of space servers and their associated cabling and electronics occupy in the data centre, blade servers can help cut down on data centre costs. Measuring in at around one and three-quarter inches wide, up to 20 such servers can fit into one ‘enclosure’. (A typical server rack contains 14 horizontal enclosures.)
Power consumption: A greater number of servers can be contained in the same chassis, sharing cooling, power and cabling resources. This means the cost and provision of these basic resources can be spread over a greater number of systems.
Cost: Although their initial cost is slightly greater than typical rack-mounted servers, vendors claim that, in the long-term, total cost of ownership is reduced because of easier management and space savings. Depending on the vendor and the sophistication of the product, blades tend to sell for around €100,000 to €300,000. (The cost of mid-range to high-end conventional servers can run into the millions).
Management: Most of the established vendors claim their blade products will easily hook into organisations’ existing systems management software consoles. Some of the newer vendors offer sophisticated combinations of servers and software, but with limited interoperability with products from other vendors.
A dearth of standards is holding back the adoption of blade servers. Many of the early iterations have been based on proprietary technologies, a situation that raises the threat of vendor lock-in for early-adopters.
The key standards issue lies in how the servers connect to each other and to their corresponding storage. The standard PCI (peripheral component interconnect) connector architecture used by personal computers cannot support fast exchange of data – the commercial environment that blades were designed to operate in.
Some new standards are emerging, but vendors have failed to agree on which one will ultimately win out. HP, for example, has standardised on CompaqPCI, based on technology Compaq acquired from a company called Ziatech in 2000. Other vendors believe Infiniband, an ultra-fast switched fabric connector technology, will be the most likely technology on which blades will standardise, but the technology is still immature (see also Getting up to speed). Chip giant Intel is also working on developing a chip-to-chip connector technology code-named Arapahoe, which is due for release in 2003.
First, blade servers’ intelligent allocation of resources could contribute to so-called self-healing, or ‘autonomic’ computing, where systems automatically diagnose and resolve problems – like the human body reacting to pain or illness. As blade server technology matures, much of the diagnostic work usually performed by systems administrators could be performed automatically.
Second, blade servers are likely to figure in grid computing. This is where huge numbers of computers dynamically share processing resources across private networks, and in some early instances, the Internet. Much of this work – often processor-hungry applications such as scientific research – is currently done across high-end Unix or even mainframe systems. As these organisations become more constrained for space, however, it is conceivable that they could run these grid-optimised applications across large clusters of blades.
Blade servers condense the maximum amount of computing power into the smallest possible space. Their tiny footprint, combined with sophisticated management software, makes them an attractive proposition for organisations looking to either make their server operations more efficient or increase server power at lower cost.
However, most organisations that have implemented blades to date have done so only as an add-on to their existing server infrastructure, running simple applications such as web page serving or caching. If adoption of blade servers is to increase at a quicker pace, vendors must iron out the standards and interoperability issues surrounding how they connect to organisations’ existing architectures.
Once these issues have been addressed, blade servers will become an important feature in the enterprise computing market. But it seems unlikely that they will replace the high-end, high-power ‘workhorse’ servers that process organisations’ core applications.