Trends > What's New

Presented by:


PENTAX - Continuous Laser Printers

The Blade-Mainframe Polarization Trend

The computing marketplace is becoming polarized. IT buyers are gravitating toward mainframe/scale-up systems at the high end and blade systems as mid-range and low-end server replacements. As evidence, last year IBM reported mainframes grew 8 percent in a market that was supposed to be flat; blades are growing at a 50- to 75-percent compound annual growth rate according to IDC. Sales of mid-range systems in the middle are flattening out.

The reasons for this buying trend are similar. Both blade and mainframe architectures focus strongly on:

• Increasing system utilization through virtualization
• Internal networking to improve data delivery speed
• Densely packing computing power that shares resources such as power supplies and network facilities in order to help overcome real estate/space issues
• Reducing energy consumption
• Reducing management costs through consolidation, virtualization and provisioning

Conversely, using a traditional approach based on distributed systems architecture is fraught with pitfalls:

• Systems are typically configured to be used at 10 to 20 percent of capacity (in order to leave headroom for processing at peak intervals). This means that 80 percent of the potential computing power of a distributed computing environment is wasted (along with the energy needed to power lightly-used computers).
• Networking is external between systems (rather than internal between processors and I/O) meaning that IT buyers need to purchase and power multiple (tens, hundreds or thousands of) network switching/routing devices in order for systems to share information and data.
• Distributed systems each have their own power supply, their own network interface card (or cards), and so on. In other words, they don’t leverage common systems components.
• Distributed systems burn approximately the same amount of energy all day long regardless of whether they’re being used or not. Scale-up systems, by contrast, burn more or less constant amounts of energy, but because their utilization rate is higher, less energy is wasted per compute unit delivered. And blades, by contrast, also have high utilization rates and can shut down unused, power-hungry processors until demand for their computing resources are needed.
• It simply takes more people to manage more systems than it does to manage fewer systems. The fewer people that are needed to administrate, secure and otherwise manage a computer system, the less the ultimate management cost. (Salaries for systems/network/database/storage administrators are a huge expense in every datacenter.)

Next page: >>