RBoard 750G Bandwidth Test

tombee79 · June 16, 2010, 7:31am

Is this normal for this device? I thought i will get at least 50 Mega Bits Up and 50 Mega Bits Down, Doing bandwidth testing Up/down at the same time. And i get around 18-21 Megs up and down.

I am cable connected to Eathernet port 4 on the RBoard,

When connected thru Accuss WL-520UG access point i get arount 13 down 7 up? Why is so LOW.

Thanks

If i bridge few AP with this router my bandwidth will drop to something like 1 or 2 MB for end user eth the end. SUCKS.

Thanks for you explanation.

macgaiver · June 16, 2010, 7:56am

When testing router throughput performance you need to test bandwidth trough the router (not to or from the router) - so router needs to be middle point.

You can use 24h free demo license on some X86 box and run test from your laptop to this X86 with RouterBoard in middle.

tombee79 · June 18, 2010, 5:40am

Yea i tested like you told me with MT in the middle and similar results. Only DHCP connectes those 2 PC.

I got approx 6-8 MgB up/down avrage.

What is wrong?

Thanks for any sugguestions

Chupaka · June 22, 2010, 6:45pm

you mean, 6-8 Mbps? what is CPU load of RB? I have just installed RB450G, it’s doing NAT (netmap), a bit of IPIP tunnels… and routes upto 200 Mbps

by the way, that’s it =)
http://forum.mikrotik.com/t/not-to-do-weird-instalations/28333/134

tombee79 · June 24, 2010, 2:03am

THAT IS NOT POSSIBLE 200mbs . You must miscalculate something. I got like approx 15/16 meg down/up. I AM REALY PISSED OFF. MIKROTIK SUCKS. MY 10 year old SOHO router is faster with up/down i gives 50 meg up/down.

fewi · June 24, 2010, 2:45am

You’re doing it wrong. You have something severely misconfigured.

RB750g, iperf test with default settings between two routed interfaces comes up with 171mbps:

[admin@rb750g] > /system routerboard pri
       routerboard: yes
             model: "750G"
     serial-number: "219601AD4A2F"
  current-firmware: "2.26"
  upgrade-firmware: "2.26"

[admin@rb750g] > /ip address print where address~"^192"
Flags: X - disabled, I - invalid, D - dynamic 
 #   ADDRESS            NETWORK         BROADCAST       INTERFACE                                                                                                       
 0   192.168.1.1/24     192.168.1.0     192.168.1.255   home1                                                                                                           
 1   192.168.2.1/24     192.168.2.0     192.168.2.255   home3            

sh-3.2$ ifconfig en0 | grep 192
	inet 192.168.2.254 netmask 0xffffff00 broadcast 192.168.2.255
sh-3.2$ ./iperf -p 10000 -s
------------------------------------------------------------
Server listening on TCP port 10000
TCP window size:   256 KByte (default)
------------------------------------------------------------
[  4] local 192.168.2.254 port 10000 connected with 192.168.1.252 port 59440
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-30.0 sec    611 MBytes    171 Mbits/sec

Chupaka · June 24, 2010, 6:47am

huh… look at the daily graphs =)

ether1:

CPU:

tombee79 · June 24, 2010, 11:07pm

my is only around 1mb

fewi · June 24, 2010, 11:10pm

The most likely culprits are firewall rules (filters, mangle and NAT) as well as queues. How busy is the CPU at any given point in time? How much free memory do you have?

Chupaka · June 24, 2010, 11:27pm

that my router is doing netmap NAT for all traffic it serves

fewi · June 24, 2010, 11:32pm

Mine has a firewall filter set installed.
It just depends on the content. Doing layer 7 regex checks to decide what source address to NAT to would be way more expensive than a simple netmap (that’s a random example, I don’t know why anyone sane would attempt that). I don’t think there’s anything in RouterOS you can’t configure in such a suboptimal fashion that it affects the router negatively.

rmichael · June 25, 2010, 12:06am

btest 0.1 seems to have a bug:

Win7 w/btest --[udp 1500]-- Win XP w/btest maxes out 17-24Mbps, one way (just tested it)

That might explain poor test results.

tombee79 · June 25, 2010, 1:05am

michael

rmichael can u describe in words instead of abbreviations. i am new to this mikrotik
the 1 i used for testing is v0.1

Are you saying you have the same problem with the testing ? Any one else having the same issue with bandwidth Test v0.1 Say YEA?

thx

rmichael · June 25, 2010, 1:09am

btest between two computers, one running Windows7 and second running Windows XP will not go above 17-24Mbps for whatever reason. This means that btest is not a good program to test throughput with.

tombee79 · June 25, 2010, 1:11am

Good to share opinons/problems on this forum. Do you know some other tool to use instead ?

Thanks

rmichael · June 25, 2010, 1:21am

Since you’re on windows you can try jperf 2.0.0 (I 've had issues with 2.0.2 btw)

roadracer96 · June 25, 2010, 4:54am

Maybe you know something I dont, but a gig port can technically push 1000mbps. Dunno why you think it isnt possible for it to do 200mbit. Oh wait. WE are miscalculating.

I have tons of 450Gs out in the real world. Doing NAT, via web proxy, lots of stateful firewall rules, I have seen 50mbit of real internet throughput. In-house, I easily max a 100mbit switch port doing http downloads from an internal server in the same config through the firewall. Doesnt even break a sweat.

FWIW, getting 20mb total up/down bandwidth is about right for a 54mbit access point.

EDIT: Mikrotik may have its flaws, but when you come in here screaming that something doesnt work, accusing OTHERS of miscalculating and doing things wrong, it doesnt help. Guaranteed that YOU are doing something wrong or miscalculating. The shit works…

EDIT2: Ive pushed more bandwidth than that over an OpenVPN encrypted link on a 450G, IIRC…

Chupaka · June 25, 2010, 1:18pm

BTW, it’s not possible, if we speak about TCP =)

fewi · June 25, 2010, 1:24pm

I know, NAT only happens on the first packet of a connection. It was just some random example of doing something expensive that would slow the router down. So if you want it to be technically correct, too, let’s assume I meant ICMP or UDP packets that are either stateless or the first packet in a connection for a protocol that has proper payload on the first packet.

tombee79 · June 26, 2010, 2:55am

Hi

First Download Scenerio
I tested with iperf with 2 PC runing winXP LAN IP/SUBNET 192.166.0.0/24

Note ( DLink AP N B G mode for Client NICs "network interface cards "

2. Second Download Scenerio
I tested with iperf with 2 PC runing winXP LAN IP/SUBNET 192.168.55.0/24

NOTE:
I user Asus AP bridged with MT Router Board 750g. Everything disable on the Asus AP except Antennas. I have hot spot disable only default firewall rules are enabled. That is it.

Still the same results in the range of the first test? Is it because i have B mode so it cuts down to so low in up/down, or maybe windows PCs causing it?

What do you think?

P.S will test 2 PC now conected by cable directly to MT ethernet ports. See what is going to happen then.