By: Jun Alejo
Blade server computing is fast transforming the server industry with a host of advantages in terms of design, functionality and total cost of ownership. By separating CPU and memory from other components such as cabling, power supply, network connectivity and cooling systems, blade servers significantly reduce massive enterprise server architectures into highly compact and dense form factor. According to market research company IDC, the blade server market is projected to represent nearly 29 percent of server unit shipments worldwide by the end of 2008.
Although its inherent compact design enhances scalability and ease of use for IT administrators, a blade server poses design challenges with regard to the installation of direct attached storage (DAS) devices. Since several blades already share the same power supply, cooling system and chassis, the most logical thing to do is to utilize a low-power, low-heat and compact hard disk drive. The first thing that comes into one's mind is the 2.5-inch mechanical disk drive, the same disk drives installed in notebook computers. However, for enterprise-wide applications, a better solution exists in the form of solid-state disks (SSDs), in particular flash SSDs.
Flash SSDs are high-performance, rugged plug-and-play storage devices that contain no moving parts. Using flash memory chips for storage, these devices are available in the same industry-standard form factors (2.5-inch, 3.5-inch and PMC) and interfaces (Fibre Channel, SCSI, ATA/IDE) as hard disk drives, but instead use flash memory chips in lieu of rotational magnetic disks to store data. This article seeks to explain and illustrate how flash SSDs stack up against other storage devices in blade server applications.
One SSD variant in the market today is the DRAM SSD. Used mostly in large, rackmount architectures, this device cannot be considered a viable alternative in blade server storage due to several design limitations. Owing to the volatile nature of DRAM memory, a DRAM SSD requires its own power supply, cooling fan and disk backup for data retention. These components eat up valuable space inside the drive chassis, limiting total storage capacity (less room for memory chips) and form factor (smallest DRAM SSD is 3.5-inch). Consider further its weight, and you'll have a recipe for disaster in embedded systems design.
As mentioned earlier, the most important factor in finding the right DAS for blade servers is its impact on overall system performance and cost effectiveness. In terms of power consumption, mechanical hard disks typically devour around 500mA while flash SSDs consume a mere 50mA. The difference may seem insignificant in small enterprise apps, but for huge data farms, the cost savings become apparent. This further enhances the blade server's advantage over proprietary systems with regard to operational costs.
The reliable performance of mechanical disk drives can only be ensured if these drives operate within specified temperature ranges. As drive manufacturers introduce newer models featuring spindle speeds as high as 15,000 RPM, cooling has emerged as a major issue. In fact, some suppliers are providing a dedicated cooling fan or fan/heatsink combos for optimum high-speed drive performance. Unfortunately, the cooling systems of blade servers are shared, and there's no room for these add-ons.
1 | 2
If you found this page useful, bookmark and share it on: