CRS226-24G-2S+ Poor routing performance

Hardware MikroTik hardware general
1

Last week I bought a Cloud Router Switch CRS226-24G-2S+ to replace two small Cisco switches (SG-200) and my RB493G (which was running absolutely fine and trouble free over the last 4 years), because we have needed more ports and I thought it is nice to have only one device left.

After configuring the new device I did some performance tests.

The switching performance of the new CRS226-24G-2S+ is absolutely fine (indeed wire speed between all ports), but if I take one port out of the switch group and use it for routing, the routing performance is very disappointing (pls. see the figures below), even keeping in mind that this device is mainly a switch.

Thinking about that, this group of devices (Cloud Router Switch) shouldn’t be advertised as “Perfect SOHO gateway router, switch…all in one box, Ethernet, Fiber, or 4G (with optional USB modem) gateway connection to Internet” and also the naming is a little bit misleading.

And there is another confusing thing: On the bill and on the router package itself the unit is stated as: “Cloud Router Switch CRS226-24G-2S-IN with Atheros AR9344 CPU, 128MB RAM). Accordingly to the specification it should have the QCA8519 CPU and only 64MB (which is shown in ROS). Is this only a printing mistake or are there indeed different configurations available?

On the other hand, if the CRS226-24G-2S-IN has the AR9344 CPU (which is similar to the RB2011) then I don’t understand the low routing performance (which should be at least similar to the RB2011 or to my old RB493G). This poor performance was also mentioned within another thread for the device CRS125-24G-1S-RM (http://forum.mikrotik.com/t/performance-issue-with-crs125-24g-1s-rm/74755/1). Will any ROS update improve the performance in the future?

For now I will use the new CRS226-24G-2S-IN as a pure switch and continue using my old RB493G as a pure router and think about buying a CCR1009-8G-1S-1S+ in addition (which I tried to avoid) or wait for the new RB/850G RB850Gx2.

RB493G
version: 6.12
build-time: Apr/14/2014 09:27:45
free-memory: 234.5MiB
total-memory: 256.0MiB
cpu: MIPS 24Kc V7.4
cpu-count: 1
cpu-frequency: 680MHz
cpu-load: 1%
architecture-name: mipsbe
board-name: RB493G
platform: MikroTik

Downstream: 120Mbps (my ISP line speed)
Upstream: 12Mbps
NAT rules: 26
IP filter rules: 28

CRS226-24G-2S+
version: 6.12
build-time: Apr/14/2014 09:27:45
free-memory: 40.3MiB
total-memory: 64.0MiB
cpu: MIPS 24Kc V8.5
cpu-count: 1
cpu-frequency: 400MHz
cpu-load: 100%
architecture-name: mipsbe
board-name: CRS226-24G-2S+
platform: MikroTik
Downstream: 40Mbps
Upstream: 12Mbps
NAT rules: 1
IP filter rules: 0

2

CRS226-24G-2S+
version: 6.12
build-time: Apr/14/2014 09:27:45
free-memory: 40.3MiB
total-memory: 64.0MiB
cpu: MIPS 24Kc V8.5
cpu-count: 1
cpu-frequency: 400MHz
cpu-load: 100%
architecture-name: mipsbe
board-name: CRS226-24G-2S+
platform: MikroTik
Downstream: 40Mbps
Upstream: 12Mbps
NAT rules: 1
IP filter rules: 0

Do You have any updates on this poor performance routing issues on CRS226-24G-2S+?

Thanks.

Vedran

3

Have you tried 6.15? http://www.mikrotik.com/download
Maybe it improve your routing speed, but not forget CRS have only 400Mhz CPU and rb493G have 680Mhz

4

Hi,

I was planning to buy couple of this, but after checking performance reviews this puzzled me a little.
This device has 2 x 10 Gigabit SFP+ interfaces and can only achieve:

Downstream: 40 Mbps
Upstream: 12 Mbps

Isn’t something wrong here, very wrong?

Thanks.

Vedran

5

The CRS series was not intended for high speed routing.
It is a L3 SWITCH with minimal low speed routing capabilities as a side effect for having ROS for administration purposes on it.
So one can do ACLs and maybe some tunneled access to the CLI/UI, nothing more.

And regarding the use as a SOHO router: most ppl. don’t have 10G fiber at home, not even 1G.
And for a crappy DSL line, it does its job.
If one needs gigabit speeds and really wants Mikrotik, than she/he should get a real router, at least a RB1100.
In SOHO environments, a gigabit home class router with hardware NAT and a dumb gigabit switch will do a far better job than a CRS on its own.

6

Yes. You selected the wrong device.

7

This is mainly a switch. have witching performance, not Routing - But. there is something in Your setop/software make cpu goes to100%. It should not in low speed like this. What does a /tool profile say about cpu usage? I doubt its Routing. (as long as its simple Routing, and not a bounch of Routing filter etc)

8

Would be cool to see a MikroTik Switch like this, sorted with a small Firmware like SwitchOS
with layer2 and layer3 capacities, features and options like greater or bigger Switches are offering.

9

I totally agree. CRS with an expanded version of SwOS will be much better and a lot easier to configure too.

10

For sure I accept that they have not the time yet to write a totally new SwitchOS sorted
with many options or plain functions offered by the most and well known bigger and by the way
I know also that for RouterOS, license fees must be paid and for SwitchOS not.

But for sure they can also set up the switches with a layer2 or layer+ standard license
and something like 10 € on top for the layer3 license or all in all 10 € more for each switch
and than layer3 comes as the standard there.

11

Something must be wrong when a RB750G performs the following:

NAT

Downstream: 780 Mbps
Upstream: 775 Mbps

Routing

Downstream: 997 Mbps
Upstream: 997 Mbps

12

You are forgetting, that CRS226 has 400Mhz CPU, RB750G (also RB2011 & RB493G as he is mentioning) has 680 Mhz CPU.

One is meant for routing, ather one is Switch with additional L3 capabilities
(Also dont forget that most L3 switches doesnt do NAT at all, so MikroTik clearly has a bit of advantage for SOHO powerusers)

For example, my CRS125 @ 750Mhz will do approx ~350Mbps with NAT and simple 15 Line Firewall config.

Also you should probably know, that CPU load VS. Traffic troughput, is not linear on network devices, it depends from many factors…

13

RB750G at 100MHz

Downstream: 150 Mbps
Upstream: 145 Mbps

RB750G at 150MHz

Downstream: 200 Mbps
Upstream: 195 Mbps

RB750G at 200MHz

Downstream: 250 Mbps
Upstream: 245 Mbps

RB750G at 300MHz

Downstream: 380 Mbps
Upstream: 375 Mbps

RB750G at 400MHz

Downstream: 500 Mbps
Upstream: 495 Mbps

RB750G at 600MHz

Downstream: 700 Mbps
Upstream: 695 Mbps

RB750G at 680MHz

Downstream: 780 Mbps
Upstream: 775 Mbps

RB750G at 800MHz

Downstream: 950 Mbps
Upstream: 945 Mbps

14

WOW…thanks, that is a really useful information :smiley:

Which now makes me wonder, why CRS is approx 50% slower on the same Clock speed…
(I at least hope that there are technical reasons, not business reasons behind this) :unamused:

15

It was approximately, I rounded a bit up and down to the nearest ten megabits per second. The official tests were done at 400 MHz, take a look at the following page.

http://routerboard.com/RB750G



Only 50% slower, JeLi claimed 40 Mbps.

16

Compared to my results, unfortunately I dont have a CRS226 to play with, only CRS125 so I have no way to compare them… :frowning:

17

After a couple of weeks and upgrading the CRS226 device to the new OS version 6.17 and the board to the firmware 3.17 I repeated the (routing)performance test of this switch:

Downstream: 25Mbps (120Mbit/s ISP line speed downstream)
Upstream: 12Mbps (my ISP line speed upstream)
NAT rules: 1 (only the “standard” masquerade rule)
IP filter rules: 0
cpu-load: 100%

In the meantime I have bought the CCR1009 router for routing/firewall purposes and I am happy with this device.

The CRS226 is fine for SoHo environments with a ISP line speed not more than 20Mbit/s. For everything above a proper router is necessary.

18

JeLi, How many firewall rules do you have in the crs226?

19

As I wrote above already, there is no Firewall rule: IP filter rules: 0.

20

Do you have new information?
Here is my experience with CRS125, CRS109 and RB2011.

Installation: home network (standard NAT and few FW rules, road warrior vpn usually not active)

CRS125-24G-1S-RM v6.27 (the same 6.25, 6.26)
Downstream: 30-120Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

RB2011UiAS-2HnD-IN v6.25 (on other Place)
Downstream: 200Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

i thnk, OK CRS125 is a Switch? But then i test CRS109-8G-1S-2HnD-IN

CRS109-8G-1S-2HnD-IN v6.15 (out of the Box)
Downstream: 200Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

CRS109-8G-1S-2HnD-IN v6.15 (Export/Import Config from CRS125-24G)
Downstream: 200Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

CRS109-8G-1S-2HnD-IN v6.27 (Export/Import Config from CRS125-24G)
Downstream: 120Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

WTF!? then it test…
CRS125-24G-1S-RM v6.15
Downstream: 200Mbps (200Mbit/s ISP line speed downstream)
Upstream: 20Mbps (20Mbit/s ISP line speed upstream)
cpu-load on Speedtest max 80%

I’m sure somewhere after v6.15 is a mistake somewhere!

Does anyone have similar experience or have any idea what is the problem?

EDIT: i open a neu post http://forum.mikrotik.com/viewtopic.php?f=3&t=94822