RB750r2 tunnels performance tests

RouterOS General
1

The objective of this testing is to try different scenarios and test the performance of the Mikrotik RB750r2 router to establish point to point conections solely on the network perfomance, ignoring security, privacy and other specific consdierations (broadcast domains, multi spoke vpns, etc)

To execute this test I’ve used 2 Mirkotik RB750r2 routers with the current ROS 6.38.1 connected to a 10/100/1000 switch. These routers have 100mbit interfaces.

The network diagram is the following. The objective of the test is to force a minimum internal routing and not meassure merely the traffic generated by the routers themselves between their WAN ifaces.

The test compares the performance of both TCP and UDP originated from the computer with the LAN interface of the destination router. The configuration references used for each test are linked in each.

Router A (Destination)

  • WAN: 192.168.100.170
    LAN 10.0.1.1

Router B (Origin)

  • WAN 192.168.100.71
    LAN 10.0.2.1

The first test establishes a baseline for the performance. I set up the routes to the LAN segments via the adjacent routers, so that the 10.0.2.0/24 LAN clients can access the clients located in the 10.0.1.0/24 and backwards.

Routes:

  • Router B: 10.0.1.0/24 vía 192.168.100.170
    Router A 10.0.2.0/24 vía 192.168.100.71

This test just stresses the routing capacity of the router yielding a reazonable result at about 96Mbps sustained rate.

For the next test I eliminate the static routing and establish an IPSEC session between the WAN interfaces of each router according to the following document.

http://wiki.mikrotik.com/wiki/Manual:IP/IPsec#Site_to_Site_IpSec_Tunnel

Test with IPSEC tunnel between router A (Origin) and router B (Destination). BTTest client running on the PC connected to the router B (Destination). The test results show a performance with UDP of less than 30mbps, with an average of 27mbps. TCP averages at 24mbps with a max rate of 25.3mbps. This test also shows fluctuations on the transfer rates, droping even to less than 10mbps at times.

For Test number 3 I’ve established a PPTP tunnel. Note that PPTP tunnels have a minimum and flawed security, widely known and easyly exploited. The configuration was set according to this document:

http://wiki.mikrotik.com/wiki/Manual:Interface/PPTP

The PPTP tunnel performance spikes at 88mbps with 84mbps average for UDP. The TCP results show an average of 50mbps.

For the last test I’ve defined an IPIP tunnel, that basically encapsulates IP packets inside another IP packed (limited to IP, not like GRE). The setup was done according to this doc:

http://www.mikrotik.com/testdocs/ros/2.9/interface/ipip.php

The IPIP performance showed peaks at 95mbps and was very stable at 94.5 mbps for UDP. For TCP the average was 88mbps with 90mbps spikes.

I’ll do this same test with a CCR router in about 15/20 days. I’ll also try to stress with multiple tunnels to the same Cloud Core Router.

2

You should not be running BTest server on any of the tunnel end-points itself. BTest is known to be CPU-hungry, so running it on the tunnel end-point very likely impacts the final results of your tests.

3

Cool.
If someone can make same tests with rb750gr3 is will be great.

4

Thanks Andriys. I agree with that. But running all the tests with the same format the impact should be identical to all of them.

5

That’s not correct for at least two reasons. First, not all of your tests are CPU-bound. And second, even for CPU-bound tests the performance degradation due to BTest running on the end-point is not linear in general.

6

If you forward traffic over the router, then 750r2 can get up to 40Mbps of ipsec traffic, r3 can get ~450Mbps with 1400byte packets due to HW acceleration.

7

Beware there is an issue with CCR hardware encryption, so you may be wasting your time on that until they fix it.
See: http://forum.mikrotik.com/t/is-re-ordering-fixed-yet-with-ipsec-and-hardware-acceleration-updating-thread/101814/1