Low throughput via CRS112-8P-4S

I have this device in routing mode.

admin@Switch_Hv] > ip address print
Flags: X - disabled, I - invalid, D - dynamic

ADDRESS NETWORK INTERFACE

0 X ;;; defconf
192.168.88.1/24 192.168.88.0 bridge
1 10.100.8.1/28 10.100.8.0 bridge
2 10.100.4.129/26 10.100.4.128 ether5
3 10.100.5.129/26 10.100.5.128 ether3
4 10.100.0.3/25 10.100.0.0 sfp9
5 10.100.247.1/24 10.100.247.0 ether4
6 10.100.9.249/29 10.100.9.248 ether6

There is an optic uplink but max throuhput is only over 30Mb/s because cpu in switch is at 100% and I dont know why. Theoretically this cpu has 94.7Mb/s in routing mode wih 25 ip filter rules (I havent any rules) so I think 30Mb/s is not ok.
I tried to upgrade latest 6.49.6 with no change.
There is about 30-40% cpu usage with very low traffic (kb/s).
This is usage with 3Mb/s traffic

admin@Switch_Hv] > tool profile
NAME CPU USAGE
www 0%
spi 3.5%
ethernet 2%
console 2%
graphing 0%
networking 3.5%
winbox 4.5%
logging 0%
management 13%
routing 0%
profiling 1%
telnet 0%
bridging 0%
unclassified 3%
total 32.5%

Is this limit of this model/cpu and I have to buy another device? There is a problem MT has not many useful models (maybe netPower 16P only, there is better cpu)

This device is a switch … routing requires CPU interaction and device’s CPU is pretty slow one. So while device can route, it can’t do it with any decent speed.

Mind that having winbox connected to device does hit it pretty hard (see the 13% of CPU for management process) so you’d better not have winbox connection to device open all the time.

So yes, you’ll have to purchase a router … which exact model depends on requirements.

There is a problem MT has not many useful models

Actually there are plenty of really good Routers ( in your case ) to choose …

…with full-size Ethernet frames, which occurs only during ideal traffic flow. Real flows include partial frames, so you move a column over in the test results and find 32.8 Mbit/sec, matching more with your results.

Obviously your real-world flow isn’t precisely one or the other, but a mix. However, budgeting based on the 512 byte column rather than the 1518 byte column is wise.


MT has not many useful models (maybe netPower 16P only, there is better cpu)

I think what you mean to say is that MikroTik has few models with many active PoE outputs, and they divide into wide jumps: hEX PoE to CRS112 to netPower 16P to CRS328-24P. If so, I agree that it would be nice to have that space more finely-divided, to have more options. The one I’m waiting for is basically the CRS version of the new CSS610-8P.

In the meantime, it’s standard practice to use a switch and a router: let the router route and the switch switch. This gets you many PoE outputs at wire speed coupled with fast routing.

The only thing that has me concerned is that you’ve got a separate subnet per port, so in fact you don’t have any LAN switching going on between ports at all. That leads me to recommend the new PoE version of the RB5009, which is a router with switching capabilities, as opposed to a switch with routing capabilities.

As wide a gap as there is between the old CRS1xx class design and the modern switch chip in the RB5009 series, it’s possible the RB5009 is a better switch than the CRS112, too. If not, then at least you have a CRS112 to offload some of the traffic.

Even this is solvable using VLANs … each switch port is made access port to separate VLAN, one of ports is used as trunk port towards router. Router is then a typical “router on a stick” routing traffic between VLANs.

never use mikrotik switch as routing, use it as switching only and vlan that’s it, i have mikrotik switch also and use only for switching work fine but, if you try to use routing then performance will be drop drastically. remember mikrotik its cheap product use wisely.

RB5009UPr+S+IN looks very good…if Ill be able tu put in to the case Im in.
Dimensions are on the edge.