When choosing a switch for a network, I’ve noticed that the decision is rarely just about the numbers on the specification sheet.
Port count, PoE budget, uplink speeds, VLAN support, and switching capacity are obviously important. But there’s also the question of how well the switch fits into the rest of the network.
For example, if a network is already built around MikroTik routers and RouterOS, does that make a MikroTik switch the more practical choice even when another switch might offer better specifications for a similar price?
I’m curious how others approach this in real deployments.
When choosing a switch, how would you rank these factors?
Port count and PoE requirements
Uplink speed
VLAN and Layer 3 capabilities
Management and configuration options
Compatibility with the existing network
Reliability and long-term support
Price and future expansion
Would you generally stay within the same ecosystem for easier management and consistency, or choose the switch purely based on its individual capabilities?
And for those using MikroTik switches in production, has the wider RouterOS ecosystem actually made your network easier to manage?
I personally would choose the switches/routers based on what i need (so i guess on individual capabilities).
The answer to everything stated here is a big fat depends.
If you're building out a network for a huge corporation, money is not an issue, you get to make a plan and adjust accordingly with their departments depending on the budget.
Port count and PoE requirements, port count; aka port density of the switches heavly depends on what you're using the switch for and in which layer are we talking about if we are talking about a distribution switch you're gonna need a 48 port switch with some sfp(+ or 28) cages to uplink to a ToR switch. If its just a regular edge switch, 24 port might be just fine..
PoE depends if you want to use it to power accesspoints or other devices (U might need PoE, PoE+, PoE++, PoE+++), all that alone affects the price differently.
Uplink speed i tend to go for 25/40 gigabit if its ToR switch if its not 10 gig will be enough (again depends on what you use it for and where for what purpose), if its an aggregation switch i would go for a 100 gig breakout.
VLAN and layer 3 capabilities, on some locations they don't even use those, so depends if u need them.
For the ecosystem i say stay away, cater your network as an engineer in an engineering point of view, it has to work it has a job to do, it has to have redundancy, it doesn't have to have fancy graphs or look good while its doing it or have nice UI and cost 800€ baseline + a monthly subscription to use the damn thing (cisco, extreme networks and other corpo oriented trash...).
Personally i don't like the Controller -> adopt into ecosystem, way of where networking equipment is going and i don't see a point in that, its all just to have you locked down to a vendor and have an eye-candy, graph filled, apple feeling enduser experience and we know who I'm talking about.
This plays right back into the OP's questions, though. One of the things you get with RouterOS is a single way of configuring everything from the smallest CPE router to big monster switches to APs to media converters…
Leaning on fundamentals is excellent in principle, to know that a thing can be done despite being buried six levels deep in a foreign menu hierarchy, but there's value in not having to switch mental gears all the time. Add enough vendors to the mix, and an admin will be constantly having to rediscover how this particular device does a thing.
I do think that there should be a control plane for AP's and CAPsMAN does exactly that and that i can get behind because if u do have any serious deployment your AP's can range from different models to 3 or 50 of them.
So for something like that i can get behind a controller, but a unify definition of "controller" i don't like.
As I have said elsewhere, I let switches be switches, and routers be routers. In a lot of large deployments, they aren't even managed by the same team, so much commonality between the two is of far lesser value than within those two categories.
So, to that end, minimizing vendor/OS platforms within type has value. In larger deployments, so does limiting the number of models deployed (sparing and all that . . .) which might at times have a larger device deployed than needed (but will have a spare on hand vs. buying multiples of something new, and logistics.)
I avoid management platforms that effect vendor lock-in unless there is a stupidly good reason.
Interop isn't a concern for me, since IP networking is a standard. Anything offered that is NOT standard is rejected.
Features are a bjg "it depends" . . . Myself, I have no use for L3 in a switch in large deployments, but smaller ones might.
Support and lifespan are crucial. Vendors who love to drive revenue by way of forced upgrades due to short lifespans and EOS/EOL are not on my list as well. Vendor lifetime and stability is also key - no matter how good something is, if the mfg. is "teetering on the edge" of folding, I don't want it.
The company network infrastructure I mange at work has 26 different switches. All from Mikrotik, different types, most running RouterOS, SwitchOS on some smaller ones, CCR's CRS, CSS plus some Wireless stuff and a CHR.
Being a (almost) pure Mikrotik shop makes life much easier:
Firmware updates: The Longterm RouterOS images for ARM and MIPSBE cover all RouterOS devices except for two. I've still three larger DLINK switches left, but these haven't had updates for at least 8 years.
SNMP monitoring scripts for all switches - one set for all of them.
Spare devices in storage if a critical devices dies.
And then the political part. My employer is a European company serving multiple govermental agencies. I am supposed to buy European if possible. We once called the US and Russia our friends, but nowadays with these friends we don't need enemies. And I have seen forgotten service account or backdoors in Cisco gear more then once. I can't tell wether it was sloppiness or the influence of some three-letter agency.
That makes sense. I agree that the actual requirements should come before the ecosystem, especially when redundancy and performance are the priorities.
I’m curious though for smaller networks where the same vendor is already being used for routers and switches, do you still prefer mixing vendors, or do you think the simpler management of staying within one ecosystem can be worth it?