And the reasons for buying new IT gear are as follows ...

When patches stop being issued for your switch, for one

Sysadmin blog Working with small businesses is a master's degree in the real lifespan of IT infrastructure. Companies with real budgets replace everything regularly; refresh cycles are two or three years apart, rarely five. The same doesn't necessarily apply to SMBs, and there are very real lessons to be learned.

Perhaps the most important is that hardware lasts a lot longer than vendors say it will. Oh, some hardware – like hard drives – are functionally consumables. Hard drives that last more than about five years are remarkable exceptions and SSDs don't seem to be living much longer.

Fans are probably the most likely component to die. CPU fans, chassis fans, power supply fans, all seem to start making that awkward grindy noise around the four-year mark and around the six-year mark most just keel over and die. So on and so forth through all the bits that make up our datacenters.

Many SMBs become experts in getting 10 years out of their servers, and more out of other equipment such as switches or printers. Even after these devices have served their time on the front line in the data centre, they might be moved out to a less critical role on the edges of the network, or sent home with an employee.

Obsolescence forces churn

I have recently rewired my home network. During this exercise I had thought that I might move from my flat, VLAN-less network to something a little more interesting, and for the first time in forever I actually looked at the model of switch I was using.

To my amazement, I had been trundling along on a positively ancient pair of D-Link DGS-1216T switches for at least the past year. (My previous 8-port D-Link finally proving to have inadequate port count.)

One of these DGS-1216T is new enough to be able to do VLANs, but the other is so old that not only will it not do VLANs, I cannot upgrade the firmware to the newer version. That makes this switch at least 10 years old, if not older.

I know how I came by this switch. One of my clients uses various and sundry D-Link switches and has for 15 some-odd years. Over time, the 12 and even 24 port switches have proven to be too small and we've slowly moved them towards 48-port models. In an attempt to encourage device churn in the hopes of eventually getting switches with some semblance of modernity I have been "disappearing" the older ones.

The old switches aren't all gone. What's more, the 48-port models have their own issues: as network demands grow, the need for trunking between switches grows. Unfortunately, not all of the D-Link 48-port models support 802.3ad Link Aggregation Control Protocol (LACP), let alone 802.1aq or 802.1AX!

This not only makes mixing and matching D-Link switches from different generations hard, but it can make mixing and matching switches from multiple vendors nearly impossible.

This customer is facing a refresh. I want to move them to Scale Computing hyperconverged boxes. This isn't going to be doable unless I can do both VLANs and Trunking between all the switches on the network.

Obsolesce is the traditional trap of the SMB tendency to keep hardware in service for as long as it is technically functional. Eventually, you just need to upgrade. Functionality, however, should not be the only concern driving device replacement.

Picking a dance partner

Recent research has taught me a few things. The first is that there is a wide feature gap between switch vendors who traditionally target the SMB (such as D-Link) and Supermicro. They play in completely different worlds.

The gap between Supermicro and Cisco, however, isn't nearly so big as I had originally thought. The price gap is certainly there, but the features gap isn't quite the unbridgeable chasm I used to imagine. The big differences seem to lie in support, and the fact that Cisco's range is much wider than Supermicro's, so it is able to go beyond top of rack switching.

This is amazing, seductive ... and a problem. If vendors such as Supermicro are catching up to Cisco and offering reasonable alternatives that can meet the needs of SMBs today and for the foreseeable future, then SMBs are going to buy Supermicro and run that equipment until it dies.

With Supermicro, you start off with enough features and support for industry standards that there is every chance it could continue to serve for a decade or more. How many years will those devices be in service past their supported lifespan?

The internet of auuugh

Increased standardisation and commoditisation of functionality are double edged swords. We get far more bang for our buck, but without rapid continual obsolescence systems administrators have a new battle to fight.

Fifteen years ago baseband management controllers were very expensive additions to servers. Now they're a standard inclusion even on SMB servers. Fifteen years ago most systems administrators thought eggshell security was OK. Today we know that's completely crazy.

Dragging along old gear isn't OK. The instant the vendor stops issuing patches for your switch, or your server's baseband management controller or whatever else you're using, its time to work on replacements, even if the equipment is still technically functional.

How do you convince small businesses to throw away switches that are capable of meeting the functional needs of the business, but you want gone because of vague concerns about security vulnerabilities that haven't yet actually proven to be a problem? That's the real issue. ®