IBM x336 packs power in a slim server framework

If you are looking for lots of server power packed into a small space, look no further than the IBM xServer 336 series. In our tests, we were impressed with its solid performance, a great physical design and management features.

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.

If you are looking for lots of server power packed into a small space, look no further than the IBM xServer 336 series. In our tests, we were impressed with its solid performance, a great physical design and management features.

The server's single-rack space form will be attractive for enterprise applications where two-processor servers are needed yet space is a premium. For example, businesses with many CPU-intensive Web and database applications that require dedicated servers would benefit from the x336.

The server packs two 3.6-GHz processors, two PCI-X slots (we had one 64-bit 133-MHz slot and one 64-bit 100-MHz slot), two Gigabit ports and two 3.5-inch drive slots in one 1.75-inch vertical rack space. The server we tested also included 4G bytes of 400-MHz DDR2 synchronous dynamic DRAM.

In the test, we could saturate the dual Gigabit Ethernet interfaces with 1.96G bit/sec of traffic. The server performed well, supporting nearly 760 SSL transactions per second. In our I/O subsystem test, the x336 could support 102 transactions per second, with an average disk queue length greater than five. Comparing this with other servers, our 1-year-old server could sustain only 45 transactions per second. (Note: Results are for relative performance only, not to determine absolute server load capacity. Our results can show whether one server is more powerful than another. They can't show how many users can be serviced, since the test methodology may not simulate the application used.)

Finding errors quickly

We were impressed with the system's "light path diagnostic" troubleshooting aid, which made fixing failed components fast and easy. When the retractable operator information panel extends from the chassis, LEDs become visible, indicating whether the server subsystem is the source of a system error. A "remind" button lets the user acknowledge the system error, which clears the error LEDs and causes the system-error LED on the front panel to blink every two seconds until the error is cleared. If a new error occurs, the front system-error LED panel lights up. The error LEDs indicate individual components in an error state. For example, if a RAM module fails, the system-error LED lights. After you extend the information panel, the MEM error LED light is visible. After opening the chassis, a lit LED beside the failed RAM module indicates the problem.

Under the hood

After opening the large removable cover (the cover is easy to remove, but it takes some effort to close, as it tends to get hung up in the fan access doors), the system's processors, RAM, PCI-X slots and component error LEDs are revealed. The components were neatly laid out, with no cables to route or get bound in the chassis cover. Two hinged doors on the top of the server provide access to redundant fans, which can be hot swapped in case of failure. Early dual-processor single-rack space servers (regardless of vendor) had a cooling problem, and it appears that IBM has done a good job in addressing this issue.