For more than eight years I have been working with designers and product managers in choosing, developing and managing PC hardware platforms that get built into larger solutions. These often take the form of embedded hardware, integrated stand-alone computing platforms or appliance-style hardware running an embedded single-purpose application.
It is surprisingly common for the designers of these solutions to run into some problems due to lack of planning or misguided expectations concerning several aspects of the design not directly related to the features and capabilities of the PC hardware itself. These three common "gotchas" so often show up too late to affordably get past and remain on deadline:
Planning for the PC form factor that best fits your requirement.
Considering the reliability options available to the solution.
Taking into account the long-term consistent supply of the PC hardware.
An often overlooked yet critical aspect for choosing a system is taking into account where the unit(s) will be installed. If the PC will sit on or under a desk in a typical working environment, nearly any standard form factor will get the job done. If embedding the PC into a chassis for a larger machine, in tight spaces or with specific mounting requirements, then the form factor plays a major role in the decision process. Common Off The Shelf (COTS) computer enclosures can vary widely, but the form factor you choose also limits the capabilities of the unit, including the performance of the CPU, amount of RAM, number and size of the hard drives, number and size of PCIe expansion slots, quantity of I/O ports and the size and number of the power supplies.
I cannot begin to tell you how many times I have seen design engineers get very frustrated with their lack of choices in embedded PC systems due to the fact they designed the enclosure destined to be filled with an embedded PC without planning ahead for the required system. I will get a list of very specific features which often cannot be provided in the physical or reliability limitations predetermined by the chassis accepting the PC. In those cases, the requirements need to be altered, the chassis needs to be modified or a system must be designed from scratch and built by an ODM all of which are usually considered far from ideal.
Sometimes designers don't know about or forget about the reliability features offered in many PC products. While not the absolute rule, it is generally accepted that highly engineered, tested, qualified and mass-produced professional-class PC platforms sold by a large scale Tier-1 manufacturer are going to be more reliable over the long haul to similar platforms assembled from bits and pieces at the same price point. But does that mean the only way to achieve high reliability is to buy from one of the handful of top-tier PC manufacturers?
Probably not, but achieving the same reliability achieved by the top systems manufacturers could require significant testing and development resources and an involved QA process that can provide the level of regression and failure analysis testing that most companies not already in the PC business are willing or able to invest in. It makes sense to at least consider mass-produced PCs to ensure reliability, serviceability and troubleshooting capabilities.
Choosing the right components for a PC system can help increase reliability. Items like Solid State hard Drives (SSDs) can vastly improve the reliability of a PC during its useful life, but there are a ton of caveats with SSDs to consider. Flash memory has a known fixed write limit. After a specific number of writes to the same bit cell, the cell will fail and eventually enough will have failed for the drive to be useless. If the application your solution uses will be constantly writing to disk 24/7, then a SSD may be a less than ideal solution. In this case, the type of SSD you choose can become extremely important.
One can also choose a platform with built-in reliability enhancers that address the potential failure of certain technologies common in most PCs like power supplies, fans, traditional hard drives and other components. Options such as redundant and hot-pluggable power supplies may appear to be a simple solution, but are not extremely common, will cost more and require more physical volume to implement. There are several reliability options which are commonly offered on certain COTS platforms:
Redundant Power Supplies
Hot Plug Power Supplies
RAID-Enabled Hard Drives
Hot Plug Hard Drives
Redundant and/or Hot Plug Cooling Fans
Hardware System Management for hardware status, failure alerts and predictive failure alerts.
These features can bring better reliability when the uptime of the application running on the PC is critical to the overall solution.
When application uptime is critical, there are even more comprehensive solutions such as hardware clustering, distributed computing and virtualization which can provide up to 99.999 percent uptime. These solutions are typically very costly and require extensive development time.
Another often overlooked aspect of hardware choice is the life cycle consistency of the hardware. Once you choose a PC platform, ask yourself:
How long does that core platform need to be available for purchase?
Can your solution run on that platform if any single component changes?
What is the impact on the solution when the platform does need to change?
We have all seen the wild advancement of PC technology over the past several decades, and the pace of change hasn't slowed in recent years. A PC purchased two years ago is often considered a relic and the parts that made up that PC are likely not available today.
It is possible to get a PC platform that can be purchased today and not experience any change to that platform for up to 30 months. However, it takes a massive logistical effort to ensure the supply of parts is reliable and consistent. Be sure to discuss change management with your suppliers before committing to a PC platform.
While maintaining the supply of a consistent platform is very important during its life cycle, one must also consider what the impact of switching to a new platform when the old one is discontinued. All platforms will be discontinued at some point. Extending the availability beyond the point where the components are being manufactured can be very costly, because those parts will have to be purchased based on your forecasts and stored somewhere until needed. It is most often advantageous to move to the next generation platform when it comes out.
Planning in advance to move to a new PC platform is critical to the design process. Life cycle management is critical and should never be ignored, as many companies are caught off-guard when their platform is unexpectedly discontinued and they have to transition to a new platform.
Ultimately, a designer can get whatever they want in a PC platform if they are willing to pay any price to get it. However, maintaining the lowest possible cost is very often a high priority in most companies' solutions. If a designer can plan accordingly and not skip straight to the basic features and capabilities requirements of their PC platform, they can reduce costs, improve reliability and get the longest shipping life with the fewest changes.
For more information: