eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Scale-out architectures—clusters of smaller systems linked by a high-speed interconnect to create the computing power of larger servers—are gaining traction in enterprise data centers. As they do, OEMs are designing new servers to meet the demand.
Scale-out proponent Dell Inc. last week launched the PowerEdge 1855, its first blade system in almost three years. The launch came a week after IBM, of Armonk, N.Y., unveiled its eServer p5-575, a thin, high-end computer for cluster environments.
Dell, of Round Rock, Texas, unveiled the PowerEdge 1855, which combines a small footprint with the latest technologies and management capabilities, according to officials. Compared with Dells 1U (1.75-inch) rack servers, the 1855 gives up to 43 percent more performance per square foot, with up to 62 percent more blades than 1U systems in a standard 42U (73.5 inches) rack, officials said.
A 7U (12.25-inch) chassis can hold up to 10 servers and consumes 13 percent less power, officials said.
Powered by Intel Corp.s “Nocona” Xeon chips, which can run both 32-bit and 64-bit applications, the two-way 1855 is managed by the same OpenManage 4 software used for Dells other systems.
The new chassis also is designed to accommodate future technologies such as new processors.
The new blades are available immediately. Pricing for the chassis starts at $2,999, with each blade starting at $1,699.
Land America Financial Group Inc. has been using Dells previous blade system offering, the Pentium III-powered 1655MC, for several years.
“Theyre good from a space consolidation perspective and cost savings, but they werent a true enterprise-class server,” said Ken Meszaros, network design manager for the Richmond, Va., company. “This [1855] has everything were looking for: a lot of hot-swappable components, hard drives embedded in the system.”
Redundant power and cooling features and Fibre Channel support for storage-area network connectivity also are important, Meszaros said.
About four years ago, Land America started building a scale-up data center, buying a four-way system, but found that many software platforms performed better in a scale-out environment.
“Certain architectures just dont scale up very well, so it didnt make a lot of sense [to pursue that model],” Meszaros said.
IBM unveiled the high-end p5-575, an eight-way system designed for cluster configurations. As many as 64 of the p5-575s can be linked in a cluster called the IBM Cluster 1600. IBM previewed the system at the SC2004 supercomputer show and will make the system available next quarter, said Jim McGaughan, director of eServer strategy for IBM.
The p5-575 will run on IBMs 1.9GHz Power5 chip, supporting Linux or AIX 5L, IBMs Unix operating system. The systems initially will be linked via Gigabit Ethernet, but in the second quarter of next year, they will also support the companys pSeries High Performance Switch.
Key to enabling IBM to pack so many processors into a thin form factor is that it will disable one of the two cores on each chip, with each core having 36MB of dedicated cache memory, McGaughan said.