How we tested four servers

We tested four types of IBM servers.

An IBM blade chassis (BC-H) was used to house blades, and we mounted servers into our racks. We used a TED Model 1000 power meter that measures kilowatt-hours attached to a custom-built cabling plug and receptacle device to perform a 240VAC [volts in alternating current] split-phase measurement for the blade server chassis and a 120VAC single-phase measurement for the rackable units.

Measurements were taken directly from the equipment. We multiplied the counted kilowatts-per-hour measurement by 43,284 hours to obtain the number of kilowatt-hours the equipment would use over a five-year period including a single leap year.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

We tested four types of IBM servers.

An IBM blade chassis (BC-H) was used to house blades, and we mounted servers into our racks. We used a TED Model 1000 power meter that measures kilowatt-hours attached to a custom-built cabling plug and receptacle device to perform a 240VAC [volts in alternating current] split-phase measurement for the blade server chassis and a 120VAC single-phase measurement for the rackable units.

Measurements were taken directly from the equipment. We multiplied the counted kilowatts-per-hour measurement by 43,284 hours to obtain the number of kilowatt-hours the equipment would use over a five-year period including a single leap year.

We measured the performance of each server and blade running Windows 2003 Enterprise Server R2 and Red Hat Enterprise Linux 5 in default configurations (no optimizations were used).

The blade servers and rack servers performed essentially identically and very well.

One server, the x3650, had slightly better disk performance, we believe because of its RAID 5 configuration. Loads were spawned by scripts that simultaneously performed calculations of arrays, wrote to local storage (disks) and performed a secure FTP download.

The full specifications of the IBM servers tested are as follows:

* BC-H Blade Center blade chassis (no blades, but including a 10GB Ethernet switch, 1GB Ethernet switch, four power supplies)

* HS21 blade, Model 8853 (two 2.0GHz Intel Quad Core CPUs, 1.333GHz front side bus (FSB), 8GB DDR2 memory, two 73GB serial-attached SCSI [SAS] drives)

* HS21 XM blade, Model 7995 (two 2.33GHz Intel Quad Core CPUs, 1.333GHz FSB, 16GB DDR2 memory, one 73GB SAS drive)

* x3650 Model 7979 (two 2.66GHz Intel Quad Core CPUs, 1.333GHz FSB, 8GB DDR2 memory, Server RAID host bus adapter, four 73GB SAS drives, two redundant power supplies)

* x3550 Model 7978 (1U two 2.66GHz Intel Quad Core CPUs, 1.333GHz FSB, 8GB DDR2 memory, Server RAID host bus adapter, two 73GB SAS drives, two redundant power supplies).

< Return to main test

Read more about data center in Network World's Data Center section.