Scenario: Non-profit company with 15 users including 2 technicians looking to reduce the total cost of ownership of their computer systems.
The main goal is to free up the technicians’ time to allow them to spend more time innovating new systems and software instead of troubleshooting the old ones. Part of the total cost of ownership in this case is the opportunity cost of the lost software and product development time not just in dollars but also in moving the company strategically where it needs to go to survive. Total cost of ownership is a concept that has been the downfall of many through the ages. A great warlord was known to give elephants as gifts to his enemies. They would accept out of pride but would be brought to ruin by the cost of maintaining such a massive ego…err…creature. Instead of supporting their army or people, they would pour their wealth into maintaining this symbol of power and wealth until it was all gone.
Computers in the environment
1) Data entry and display
Expected low maintenance lifespan: 8 years (thin client), 5 years (desktop), 3 years (laptop)
2) Design and processing
Expected low maintenance lifespan: 5 years (desktop)
Expected low maintenance lifespan: 5-8 years
Computing power doubles every 18 months. Three years between upgrades means that work requiring significant computational resources will be 4x more efficient. As no computer always uses its resources to a maximal potential, this means that it takes a significant factor (~4x) of increased performance to actually impact work efficiency. If your computer was only 10% more powerful, would your productivity really go up 10%? Probably not. You would only save 10% of your time on the tasks where you are waiting for the computer to finish something before moving on. This is not a majority of the time in most usage scenarios. Therefore you would not expect to see a significant increase in productivity until there was a huge improvement in to computing that changed your computing expectations and gave you room to change your work patterns. An example could be video editing. If your final render of a 5 minute movie was 8 hours, would a 10% CPU boost change your work pattern? Not likely. You would still edit until midnight and leave it to render until breakfast. You would still only get 1-2 videos done per day. But if you had a 300% boost, you could render several videos per day. If you upgrade too often then the amount of time involved in making that transition increases your total cost of ownership without adding a whole lot of value.
Average life expectancy is 5-7 years with optimal quality and care. Build quality is important here. LCDs usually fail because of power supply or problems other than the backlight. Monitors should be set to turn off after 1 hour without use. Frequent power cycles will reduce the backlight life significantly. Buying cheap doesn’t improve your overall cost of ownership if you are constantly supporting and replacing monitors.
After doing some research (PCWorld, CNet, Lenovo.com) I’m leaning towards buying as much as possible from Lenovo (was IBM). If price must be considered first then I would recommend Dell. For our company, we ended up going with Dell over Lenovo because of the flexibility and pricing of their server configurations.
Buying from a single brand for everything should help reduce the number and types of problems that you face. I have found that supporting custom built and stock computers from a variety of manufacturers means that I have to have a wide variety of parts on hand and have to spend a lot of time tracking down really weird bugs and incompatibilities. In real life, two products from different brands that both claim to conform to the same specification and are both well rated and reviewed sometimes simply will not play well together. If you can find a brand that works for most situations, stick with that brand even if it means paying a bit more up front sometimes. You will find it easier to find the source of a problem when all of your devices start having the same issue at once or when just one of 10 is doing something weird. There will also be a large community of users just like yourself who are also or have already had the same problem and will help you solve it. If you must build custom computers, build at least 3 and make them all exactly the same.
There are no clear winners in the reliability of hard drives. Western Digital Caviar Black (5yr warranty) has been recommended often. Also, slower disks that are enterprise class, in raid 1 or 10 configurations should improve reliability. As SSDs go, in sufficient quality (Intel, Toshiba, and Samsung
) they are more reliable than traditional HDDs. New HDDs should be stress tested using Bart’s Stuff Test
or Hitachi Drive Fitness Test
. Any drive that passes either test should be reliable for 4-5 years. It is unknown how long the current SSDs will last in the long run but Intel is now offering 5yr warranties on it’s most reliable drives.
The ideal configuration for performance and reliability:
Intel X-25 SSD Raid 1
The ideal configuration for storage and reliability:
Western Digital Caviar Black HDD Raid 1
The ideal configuration for storage, performance and reliability:
Western Digital Caviar Black HDD Raid 10
Mission critical PCs after 3 yrs: Non business critical or low usage users
First problem after 5 yrs: Ebay, donation
The idea here is to spend as little time maintaining things as possible. It costs more to drop everything and fix a mission critical device. For this reason, we retire a device to a non-mission critical function long before it should start developing quirks. When it does start to get flaky, you can schedule the resolution much more conveniently because it doesn’t have to be perfect right away. However, as a device ages it will begin to consume more and more maintenance time.
After five years, there is zero patience for problems. Replace it and start over again. Drop the device on eBay, part it out, sell it to a computer shop, or recycle it depending on the condition. Cut your losses and run. You may fix it this time but something else is about to go wrong with it too. After 5 years, the tech industry doesn’t want you to own it and will not be supporting it as well. Answers will be harder to find. Your new techs won’t be as familiar with such an old system. It is all sad but true.
Centralized log monitoring
Centralized backup management
Centralized security monitoring
Avoiding time consuming rebuilds and reconfigurations involves keeping up with the warning signs of failure and having a backup available in case of failure. Automatic backups tend to be a bit unreliable so a tool that would allow a quick assessment of backup statuses across the network would be very useful. Security is always a must-have and is easier to manage from one location.
OS migration to new computers
Windows 7 appears to be very stable. Replacement computers can have the entire OS migrated to new hardware with a minimum of reconfiguration if that is desired. Software solutions that require custom software and uncommon configuration should be well documented and updated whenever changes are required. Installation media and licenses should be archived. Plugins and scripts should also be archived.
Reducing computer system maintenance time per day from X percentage (without shifting it to someone else) puts Y thousands extra (above current expenditures) into the budget per year. Obviously spending even close to that amount would result in no net improvement in a fiscal sense and would likely not translate directly into as much time savings as desired in the short run. It does, however, give some idea of the scale to keep in mind when planning and shopping.
Practices that can be traded for reliability include:
- extensive comparison shopping
- purchasing small with the intent to upgrade when needs increase
- single purchases
Purchasing in duos, or more will improve troubleshooting effectiveness, baseline performance assessment, accommodating shifting workloads, and rebounding from hardware failures.
Always cross train. Each function of the business needs to be understood in a basic way by at least two people. In the event that the employee who normally handles a job is not available, someone else can step in and fumble their way through it.