Check out the rest of our Server Month resources.
Managing a data center’s hardware lifecycle is an ongoing process in constant need of updating. Since the economic downturn, IT administrators have seen their budgets become much leaner, forcing IT to get more value from hardware investments. This usually means extending the technology refresh cycle and getting more service life from servers.
A typical server service life is about three to five years, but that number can now be as long as eight years. Some data centers are so strapped for cash that servers must run even longer than that. Machines wear down, break and often require increased maintenance as they get older. But this boosts data center costs. So the question becomes, how can data center engineers extend the life of their machines while keeping everything cost effective?
Using server virtualization to extend server life
Virtualization is often believed to be an option only for newer machines. This isn’t entirely true. There are two ways to look at virtualization. One is bare-metal architecture, where a hypervisor is installed right on top of the hardware with no OS installed in-between. This method usually requires Intel VT-ready or AMD-V-ready processors that are present in newer hardware. But in the second method, there is a hosted approach where a virtualization platform is installed on top of an already existing operating system.
Consider, for example, an older Hewlett-Packard Co. ProLiant ML330 G3 server with powerful Xeon processors and 3 GB of memory. The network card is an older 10/100 Ethernet network interface card (NIC), but the device has open PCI ports to expand. The hard drive is still solid and the components inside work fine. In most cases, this server can be classified as end of life. But with virtualization options such as the free version of VMware Server, this formerly DOA machine can be repurposed for numerous uses and operate efficiently for at least another three to four years.
Given the specifications above, this machine can easily host two or even three virtual machines (VMs) capable of running low- to medium-level workloads. By simply installing it as an application on top of a host Windows or Linux operating system, the machine will now have a thin virtualization layer that enables multiple virtual machines to simultaneously reside on a single server. This older server can be connected to a storage area network (SAN) where the VMs are stored, or simply employ a RAID group to store the virtual machines directly on the box. There are numerous real-world use cases for a late-model physical virtual host:
- IT testing and patching;
- Work with new applications and OS in a secure virtual environment;
- Snapshotting VMs and rollback as necessary;
- Re-host older legacy operating systems such as Server 2000, NT, or Server 2003.
- Provision and create new VMs to deploy later
Hardware upgrade can extend server life
Using the previous server specs for this example, adding an additional Xeon processor and more memory will certainly make a server operate longer and perform better. This machine can now be used as a virtualization host, services-based server, licensing server, or a host for nonresource-intensive applications. Numerous 1U and tower-style servers are upgradeable, but they rarely are upgraded in practice. Check the specifications of your system to verify its maximum capacity for a given component. Here are some common upgrades to consider:
- Adding a 10/100/1,000 Ethernet NIC is recommended. The server may not operate well when connected to a new ProCurve or Catalyst switch capable of gigabit network speeds. Depending on the server model, a refurbished or new NIC for an older device costs about $60 to $140.
- Adding another processor to an available socket is a bit more expensive, but it’s still preferable to replacing a server capable of working another two to three years. A replacement or extra processor with a heat sink will cost about $200 to $400.
- Memory is a great way of speeding up a machine and redistributing resources to applications still running on the server. Depending on the server’s architecture (32-bit vs. 64-bit) the amount of memory will vary. Some smaller 32-bit 1U machines can handle only 2 GB of RAM, while others with better equipped motherboards can handle up to 4 GB. The first generation 64-bit systems are designed to take 6 GB to 8 GB of memory. Check with your server’s manufacturer to determine type and sizes of memory modules that can be installed. Memory prices will vary, but expect to spend as little as $30 on additional RAM.
- Some older servers still working with 5,400 RPM hard drives offering just 40 GB of space. Data center engineers need to be aware of their machine’s capabilities as a lot of hardware can handle much larger drives than that. Upgrading to a higher capacity, higher RPM drive will improve performance and extend server life. Moving to a 7,200 RPM or even 15K RPM drive will cost $200 to $400. But if this is the only component that needs to be replaced, this is a good investment.
Best practices for extending server life
Optimizing the physical and electrical environment will help any server run longer. As you aim to extend server life, consider these factors: temperature, air quality and electricity.
- For most servers, room temperatures for normal operation should be between 60 degrees to 85 degrees Fahrenheit. Humidity should remain steady between 50 percent and 75 percent (to minimize any static buildup). Managing the environment in a data center not only increases performance of the machines inside but also allows them to operate longer.
- Air quality is also important as computers require air circulation to prevent overheating. Remember to keep good airflow around your machines. Move servers away from walls to allow vents to properly circulate hot and cold air.
- Electricity quality is a key factor in determining the longevity of a machine. Managing the amount of voltage entering the server room will be critical to keeping machines healthy. A brownout or a spike can severely damage computers. Using a good uninterruptible power supply and power regulators helps monitor the amount of voltage being pushed into a rack of servers or just one machine.
Preventive maintenance can extend server life
Older machines require extra care. At least once a month, several tasks need to be completed to help the system run longer. Make sure, for example that all operating system patches, updates, rollups and drivers have been deployed. Look beyond the operating system and ensure that antivirus system has been updated and operational.
If a server uses local storage, run the Defragment tool as well as Disk Cleanup to help the machine run longer with stable data and a cleaner running drive. Running hard drive health checks on a regular basis is a great proactive way to keep a machine healthy. Regularly check the hardware components within the machine itself. Make sure that all fans are operational and verify that all components and cable headers are in their place. Keep a careful watch on dust buildup as this can quickly disrupt cooling and cause electrical glitches that can cripple a server.
When working with older hardware, the little things can make all the difference. IT engineers are tasked with prolonging the life of their servers; now with better technology and resources, this can be accomplished more easily and cost-effectively than ever.
Check out the rest of our Server Month resources.
ABOUT THE AUTHOR: Bill Kleyman, MBA, MISM, is an avid technologist with experience in network infrastructure management. His engineering work includes large virtualization deployments as well as business network design and implementation. Currently, he is the Director of Technology at World Wide Fittings Inc., a global manufacturing firm with locations in China, Europe and the United States.