If the servers at your firm outnumber your people, you likely have way too many.  If your firm is expanding, a single server may no longer be enough.  Perhaps you would prefer to retire all your servers and run your practice through the “cloud.”  In any case, it makes sense on a periodic basis to review your server architecture to ensure it is engineered to meet the needs of the firm.

Naturally, folks in the IT field love to work with the latest, greatest and most powerful new equipment.  There is no question that new technologies can improve functionality and reduce costs.  There is, however, a clear tendency for technical folks to over-engineer and design networks that are overly complex, overly expensive to maintain, and inherently less reliable.

One of the newer technologies is called server virtualization, which can be very useful, but can also have a negative effect if not managed properly from a business standpoint. Virtualization technology allows several physical hardware servers to be consolidated onto a single physical server. These servers then become “virtual servers” or “VMs”. This strategy absolutely needs to be considered as part of any design where there are multiple servers. Unfortunately, every virtual server incurs the same software costs for licensing as if it were still a physical server. Additionally, each virtual server has largely the same concerns that a dedicated physical server has with regards to patching, monitoring, and other management activity.

Simpler means cheaper, faster, and more reliable – period. I’m not advising that you go “cheap” on hardware. You want the good stuff. Not the most expensive, but rather mainstream good technology. You also want to refresh it regularly with newer technology every 3-5 years. Unless you’re getting your people to work for free, few things will cost you as much in productivity and labor than trying to make that 5-year-old PC or server give you just one more year.  Your server and network infrastructure should be modern, but straightforward in design.

Just about every software vendor wants a dedicated server for their product.  They are afraid of applications stepping on each other and argue that if a server reboot is required, dedicated servers will impact fewer applications. That’s great in theory, but IT people often get carried away with this. The fact is that modern server hardware and operating systems are much more robust, are designed to take advantage of multiple processor cores and are engineered to avoid software conflicts.  Modern servers do not require reboots nearly as often as their predecessors.  Admittedly, I’ve had disagreements with certain technical people on this point. Speaking as a business owner, I’m encouraging a healthy skepticism on your part to help ensure that you are spending your money wisely.

There are server environments where even a few minutes of downtime (think amazon.com) can cost a fortune.  These “high availability” data centers have layers of redundancy that are very expensive to deploy.   Still, having some redundancy built into your design is worthwhile when executed properly and with restraint. I’ve seen people get carried away here to the point where complexity actually decreases stability. Too often these decisions are made without real sizing and performance data. Add to this mix hardware vendors pushing the latest, greatest server technology, network technology, even storage area networks (SANs) and you can end up with some pretty crazy designs.   

SANs are very fast and generally highly reliable arrays of disk drives.  They have their place in the architecture of many larger companies, but they belong in very few professional firms. Google doesn’t use them, because it’s not the right design for them. I’d seriously question the value of a SAN in any firm with under 1,000 people. I’ve seen software failures on SANs and the resulting cleanup and data restoration required can get downright ugly. It’s best to avoid this unless you absolutely need it.

 Modern servers are able to store very large amounts of data on built-in arrays of very fast hard drives at an excellent price point.  Solid state drive technology is even faster and more reliable; however the costs for these disk drives will likely remain too high for mainstream deployment until at least the next year.  PCs with solid state drives boot up in seconds rather than minutes if you are really feeling the “need for speed”.  Of course, the single largest concern with any server must be ensuring that the precious data it holds is backed up and secure.

If you have any questions about any of this or if you’d like a simple validation of what you’re doing with your servers and network, just send me a note at dkinsey@totalnetworks.com and I’d be happy to help.