Let's get white box servers straight

What's the difference between white box servers and Open Compute servers? What's the risk involved without brand-name server features?

There is more to modern white box servers than just a lack of brand name.

Enterprise computing is entering an age of specialization that is redefining the nature, roles and capabilities of today's data center server. All computing platforms are not created equal, which means it isn't necessary to standardize on a single ubiquitous server for every task.

A burgeoning array of server hardware balances cost, resilience and performance. High-end generalist systems from the top vendors still have a vital place for some workloads, but low-cost white box servers are gaining popularity thanks to virtualization, clustering and private cloud computing.

White box servers vs. name-brand systems

A white box server has no direct association with major name-brand systems vendors, such as Dell or HP. White box servers are made by original design manufacturers (ODMs). Distinctions generally exist between name-brand and white box systems in performance or quality, components and cost.

Although "white box" sometimes alludes to low performance or low quality, these products can provide the same computing resources as any name-brand server. The underlying processor, memory, chipset, local disk drives and other components are essentially identical and typically rely on standardized form factors and interfaces in the overall build, such as microATX motherboards, SATA disk interfaces and PCI Express expansion cards. Some major server vendors offer design enhancements or proprietary architectures to boost performance or expand manageability. ODM servers typically rely on commercial off-the-shelf components that can be assembled or upgraded for a degree of system customization.

Each white box component's quality depends on where it's sourced, which can affect the system's overall reliability. For example, the drive to lower prices might lead to a white box server with a lower-quality power supply than its name-brand equivalent. Data centers often deploy white box servers in clustered computing tasks, such as an on-demand private cloud, which effectively mitigates the risks imposed by lower-quality system components.

Similarly, white box servers are frequently deployed for non-critical computing tasks like software development and testing where unexpected downtime has no direct impact on the production environment. Instead of carving out resources on an IBM System 3950 for a media server or application development test instances, you'd use simple white box servers to save money.

White box is not equivalent to pizza box, which generally refers to any 1U or 2U rack-mounted form factor. A white box server can be a pizza box, but a pizza box is not necessarily a white box.

Virtualization features

White box configurations must properly interact with other existing systems in the data center. VM migration, as in VMware's vMotion feature, moves virtual machine instances from the memory of one server to the memory of another. VMware vSphere evaluates available processor features when creating new VMs and uses those available features to boost VM performance. This can be a problem with migration.

Moving a VM to a server with a dissimilar processor from the originating server (usually an older processor or radically different processor family) means that some features that the VM expects are not supported on the destination server, causing a crash. Enhanced vMotion Compatibility can mask some differences to protect the VM, but it's not perfect; at best it may result in reduced VM performance on systems where important processor features are absent. VMware's vMotion issues can occur just as easily between the best name-brand systems as between white box servers and white box and branded servers.

Open Compute and white box servers

The concept of vendor-agnostic white box servers overlaps the Open Compute Project initiatives for low-cost, energy-efficient servers, storage and mechanical mounting schemes.

Open Compute servers are power-efficient, barebones, high-volume, low-cost systems designed to be replaced rather than serviced in large-scale data centers. Variations of the Open Compute server design are available for AMD, ARM and Intel processors along with high-availability and system-on-chip specializations.

Open Compute designs are not necessarily intended to be categorized as white box servers; however, the Open Compute approach to server design is consistent with the simple, low-cost, high-volume computing scheme that ODM offerings trend toward.

Open Compute servers are expendable components in a virtualized, clustered and load-balanced server farm where workloads run across pools of infrastructure. This is more cost-effective for Web-scale businesses like Open Compute's founder Facebook than using high-end generalized servers with traditional computing architectures.

Should I build my own servers?

Should you go out and build your own servers? Probably not.

Most modern corporations are not in the IT business like Facebook or Google, and only need several dozen to several hundred servers to meet computing needs. It's not cost-effective to design servers from scratch and contract the build, regardless of whether you're using Open Compute designs or traditional off-the-shelf components. Most enterprises purchase generic white box servers from businesses that specialize in traditional and customizable rack designs, such as ABMX and ServersDirect.

Next Steps

Converged infrastructure vs. white box servers

Does virtualization call for a brand-name or white box server?

Will white-box switches ever be a reality in enterprise networks?

Dig Deeper on Server hardware strategy