Manage Learn to apply best practices and optimize your operations.

SAS, blades reshaping the server market

The server market continues to evolve at a rapid pace. We've identified the key server trends you need to know about.

With the year more than halfway gone, many IT managers are beginning the budgeting process for next year's IT purchases, which inevitably includes server decisions. Should you plan on upgrading your server hardware in the coming year, or should you hold off and wait for that killer technology that's just around the corner? How, exactly, do you get the most bang for your IT spending buck? To help find the answers, consider these important trends in the server space. They could help you make the most of your IT budget in the coming year.

SAS overtakes SCSI

More on servers
Blade servers: Check IT List

Open source servers serve up challenges
If you haven't heard of serial-attached SCSI (SAS) disk subsystems yet, you will soon. Disk subsystems were hitherto restricted to two simple types: Integrated Drive Electronics (IDE) and SCSI. In 1997, the average server at small and medium-sized businesses (SMBs) had practically the same architecture as a high-end desktop, an IDE subsystem. IDE soon after gave way to the SCSI subsystem, which became the across-the-board standard and has been king of the hill in the server world for the past 10 years.

SAS has entered from the bottom of the market as a cheap alternative to SCSI, but it's increasing in prevalence and exposure as speed and reliability of SAS disks improve.

Continued growth means continued complexity

Hard disks continue to grow in capacity. In 1997 the average disk was 9 GB in size. Today, you'd be hard-pressed to find a drive smaller than 72 GB. In fact, the average size of a server disk is more than 150 GB. This doesn't simplify the life of the average network administrator, though, as the more data you place onto a single spindle, the more work that spindle has to do and the larger the average disk queue length must become.

You can combat this through configuring multiple disks into a RAID configuration, using more disks to spread the information load and increase capacity. For example, a trio of 144 GB disks in a RAID5 array performs far better than a single 300 GB disk, even though the amount of available disk space is nearly identical. In addition to the performance improvements you'll realize in this configuration, you'll also enjoy the increased resiliency of a RAID configuration.

As disks grow beyond the current 500 GB high-end capacity towards 1 terabyte, it becomes more important to consider the number of disks that you deploy to balance available capacity against overall performance and fault-tolerance.

RAM ramps up

Now let's turn to trends in RAM. When many of us first started working with Windows servers, it was common to see only 16 MB of memory installed; you might splurge and put 20 MB in an Exchange or SQL server. Today, it is normal to see servers with 4 GB of RAM, a 250-fold increase in memory. During the next few years, the adoption of 64-bit applications such as Microsoft Exchange 2007 and new versions of the Windows server operating system will likely push RAM averages to 32 GB and beyond.

The sticking point is that memory pricing grows exponentially relative to gigabit size rather than linearly: a 4 GB memory module will cost more than double that of a 2 GB module, and an 8 GB module may cost more than triple that of a 4 GB module. When choosing your next server, make sure it has enough memory slots to allow you to add more RAM should you need to, even if your current needs and budget may be far less than the server's total capacity.

Blade severs making inroads

Though they've been available for several years now, blade servers are finally picking up steam in the SMB market. Previously, a small computer room might contain several servers, all using different disk types or even purchased from entirely different manufacturers. In the coming years, more SMBs will invest in server blade infrastructures connected to iSCSI SAN storage platforms. This way, when an organization needs to purchase a new server, it can simply add another blade with minimal additional investment or increase in support and maintenance costs.

Laura E. Hunter is a systems manager at the University of Pennsylvania, where she provides network planning, implementation and troubleshooting services for business units and schools within the university. Read her profile here.

Dig Deeper on Small-business infrastructure and operations

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.