Courtesy of Network Computing Just because an organization can't pay $40,000 for a single server doesn't mean it must limp along with a substandard IT infrastructure. Quality servers, software, and services are within reach--the cost of acquisition and maintenance is an ever-decreasing spiral, and workhorse 1U servers from companies like Dell and Hewlett-Packard cost as little as $1,000. In fact, more than 70 percent of SMBs use Dell and HP as their primary server vendors, according to a Gartner survey. But in our current environment of "one server, one application," you may find yourself drowning in too many servers to manage. Thankfully, you have several options for keeping your initial investment and long-term maintenance costs down and your service level high. Multicore/Virtualization. You can't have missed all the crowing about multicore servers. These X86 servers with multiple integrated CPUs are less expensive than their multiprocessor relatives but provide more processing power than single-CPU boxes. By themselves, they save time and money only if you have an application that uses too much CPU time to run on a single-CPU system, or if you have a server that handles many different tasks and needs an upgrade. In the first case, multicores can help you avoid implementing a load-balancing or clustering product But for small to medium-sized organizations, the real beauty of multicore servers is revealed when you pair them with virtualization software. For example, Citrix's and Microsoft's server virtualization products can reduce costs by making several virtual servers out of one piece of hardware. Organizations are also taking advantage of virtualization to reduce the server requirements in their data centers, thus keeping maintenance and replacement costs down. In addition, virtualization can support multiple copies of an OS on a server. This allows older applications with oddball hardware requirements to run on a virtual machine, letting you upgrade to new hardware but keep your old apps. Most of the instances we're aware of involve old DOS or OS/2 applications that count on the hardware to perform direct screen or disk writes. 64 bits. Look out: 64-bit servers are coming fast, and theoretically, they provide double the processing power in the same chip, but there's a lot more to that equation than just the CPU. The chipset and bus placed on the board with the CPU have a lot to do with the performance you'll get out of any 64-bit machine. Sure, more processing power is better. But these are not silver bullets. Most SMBs should sit tight; upgrade only if you have 64-bit applications. Blade Servers. Blade servers are the future of the data center, but be especially cautious in these early days. One big factor to consider is that most of the vendors require 208- or 220-volt power supplies. This can be a deal killer. Also look out for oversubscription both on the Ethernet network and the SAN. Many blade servers funnel a lot more through one or two ports than is acceptable to most organizations. On the plus side, many blade servers offer improved management and the ability to set up rudimentary load balancing across blades, which improves performance without increasing costs. Most vendors still charge more than for equivalent 1U boxes, but that should change soon. In fact, in a recent Network Computing poll, 56 percent respondents cited initial investment as the No. 1 barrier to blade adoption, with market immaturity second at 35 percent.
