While looking for decent 10G capable switch I found Mirkrotik has such a device for very nice price.

So I just want to ask if you can recommend this switch to connect several servers to storage. There'll be couple of optical 10G ports and about 4-5 of copper 10G.

So far we need it only as a basic switch (no special functions be employed) so maybe CRS317-1G-16S+RM is ok for this purpose. As for me, I've never tried to use pure switch solutions from MT, but maybe it is the time to try?

Beside MT's devices there are number of not too expensive 10G capable switches on the market, so what will you say?

A switch is one of those parts that can never go down when used to switch storage traffic. Otherwise you are in for a long day (well... it depends on what you are using the storage for... I assume VMs since this is mostly the case these days).

I wouldn't be very confident in using MikroTik switches for storage traffic. Especially a single switch without any redundancy.

The way I see it is that switching (especially after v6.41) is pretty much beta on Mikrotik with many enterprise features missing.
I would only consider a Mikrotik switch (as I have) for SOHO/SMB environments without any HA requirements.
Don't get me wrong... they are nice and cheap for what they do... but they are not... Cisco. As much as I don't like Cisco, their switching gear is way ahead of Mikrotik in terms of features and stability

An other alternative to 10G ethernet is Infiniband.
You can find older IB equipment at 40Gbit speeds, at a fraction of what 10GbE costs.
Of course this has its own limitations since it's not ethernet, but I found it as a nice alternative for some use cases (ie: storage traffic for a ceph cluster).

An other alternative to 10G ethernet is Infiniband.
You can find older IB equipment at 40Gbit speeds, at a fraction of what 10GbE costs.

We building some iSCSI lab network, so we dicede to not play with IP over Infiniband (but maybe we'll reconsider that).
I'd play with MT gear just for not looking for another brand for lab purpose but yes I'd like to see something that's not beta. From what I can read this switch rely on 6.41+ new bridge implementation which is neither be considered mature nor enterprise for any scale, and SwOS is kind of toy, too. All I can consider it for is its hardware (and hope for better software one day after). Hope one day it'll be able to do stacking, but no real reason for such a hope, just for the idea they'll be able to create couple of new switches before their software development team be able to use these brilliant new h/w features and they can just drop this model by that time.
MT is not Cisco, no at all, sad but still true. Small office is the place for, but it's ok for us.