In our throw-away modern society where built-in obsolescence has become routine, purchasing computer systems that last is important for businesses if they want to survive long-term in the marketplace. A business can’t simply afford to keep purchasing new systems every few years to replace older ones, even if the new ones offer all kinds of alluring bells and whistles. That may be a strategy for vendors to get rich, but it certainly won’t help your own business grow if you constantly need to replace systems and other hardware. Because I often deal with IT pros who work in various business environments, this has been a big issue in my mind for some time now. A while back I sought out advice from experts in the trenches of business IT asking them what they would suggest when your company needs to buy computers, servers, and other server hardware, and also how to make them last as long as possible. One colleague I talked with, Tracy Hardin who owns an IT consulting firm called Next Century Technologies, had some excellent suggestions on the subject of purchasing servers and I’m reproducing them here so TechGenix readers can benefit.
MITCH: Tracy, if a small or midsized business is looking to purchase some server systems and server hardware, what are some of the key features you would recommend that they keep in mind as they look around?
TRACY: I stick with proven hardware. Regardless of the brand of server, I look for these features:
- Serial-attached SCSI hard drives: SATA drives have certainly come a long way in speed and reliability, but I am sticking with SCSI. They are designed for the multiuser server environment. The performance is great, they are extremely reliable, and I rarely have to replace one. And now that the prices for SSDs have dropped we often suggest putting the OS on an SSD and the data on a SAS platter. I foresee SSDs completely overtaking the SCSI solution within the next few years.
- Hardware RAID controller: RAID is a critical part of disaster recovery, and all the servers we sell have it. SCSI drives rarely fail, but when they do, RAID will save the day in spectacular fashion.
- No less than 12GB of memory: The amount of memory will depend on how the server is utilized.
- Xeon processors: I like the Xeon line of processors because they were designed for servers.
- 3-plus warranty
- Gigabit network cards
- Set of recovery DVDs that I can leave at the customer site
MITCH: Any other suggestions regarding server hardware one should keep in mind when planning purchases?
TRACY: There are a lot of other great options if money is available, like redundant power supplies, rack-mount chassis, management cards, and hot-swap drives. In my opinion, paying extra for hot-swap drives is a waste of money. There are situations where it is a must, but given a choice, I am going to power down a server before I pull a drive, even if it’s a hot-swap.
MITCH: That’s interesting. What do you think about redundant power supplies?
TRACY: Most of the servers I sell don’t have redundant power supplies, but they all are plugged into a quality UPS with the appropriate load rating and runtime. The UPS is configured to shut down the server automatically, and, in every case, the UPS will shut itself down as well. I was called into a new client to rescue an Exchange server after an ice storm kept the office closed for five days. The brand-new UPS was never configured correctly, so repeated extended power failures crashed the server multiple times over that five-day period. Surprisingly, the server still booted, but even with Microsoft’s help, I could only recover about 80 percent of the data in the Exchange server. To this day, I configure a UPS to shut itself off after it shuts down the server, or, in the very least, configure the UPS not to bring the server up if power is restored. I also configure the UPS to email me if there is a power issue so I can notify the client that the server and UPS have been shut down.
MITCH: What about processors? Should one always go for the latest and fastest processor?
TRACY: I put more money in disk I/O and memory before I do the processors. This is a broad statement that applies to the vast majority of the servers I have sold. Virtualization and certain server application requirements can be exceptions to that rule.
MITCH: Many small and midsized business owners I talk to wonder whether it’s worth buying servers at all because of the growing ubiquity of cloud computing. What’s your opinion with regard to this?
TRACY: Why buy servers and not go “cloud”? Lots of reasons. First, the cloud is not a one-size-fits-all solution. It is simply another option. Local servers provide:
- On-site performance that can’t be matched by the cloud.
- Control over updates and maintenance schedules.
- Security: No third party has access to your data.
Most importantly, if your on-site server goes down, I can usually pinpoint the reason within 30 minutes, and no waiting on hold with someone in another state (or country), wondering what happened.
MITCH: Let’s say you decide to go with buying a server then. What can you do to make your server hardware last so you get the most mileage from your dollars?
TRACY: Some key points to making a server last six-plus years are:
- Buy more memory/disk/processor than what is needed today.
- Buy quality server hardware from a reputable manufacturer.
- Buy a UPS that is smart enough to email when there’s trouble and shut down the server.
- Keep server and UPS in a clean, cool environment that is physically secure.
- Test the UPS every year and replace batteries every two to four years.
- Keep the operating system up to date.
- Use quality antivirus/anti-malware/firewall/web filter on server and staff computers and keep it up to date.
- Use a quality business-class hardware firewall for the entire network.
- Keep users off the server console.
- Even in a clean environment, every couple of years open up the case and check for dust buildup
MITCH: Thanks, Tracy. Anything else you want to add?
TRACY: Don’t forget, all servers need a good, reliable offsite backup. Monitor backup reports and do a test restore every year. You should not have a problem keeping your server hardware running six, seven, or maybe even eight years.