Not sure if I am the only one having this issue but I have 2 bonded SFP+ ports both are 10Gb modules (Mikroitk SFP+) But all speed tests are topped out at 1Gb I am running 6.41 ROS to a 6.41.2
Is there an issue with bonded ports?
I’m using 802.3ad Mii with Layer2
Any advice would help thanks.
-Sincerely,
DesertAdmin
It seems that if you have 10Gb SFP+ and you want to have a 1Gb bottle neck just make a bond between two 10Gb SFP+ interfaces. Not sure how this is a feature.
I will being breaking the SFP+ bonds to see how those works.
-Sincerely,
DesertAdmin
I broke the bond and it is performing better but seems to be a bug or an issue with bonding and using 10GB SFP+ DACs modules.
I am running 6.41.2 on both Mirktoiks CCR 1072 both have 802.3AD with MII Layer2 on Mikrotik DACs copper pigtails 10SFP+ wave length 256nm Vend Serial: MCS16l10432
Any one else have this issue or see a solution?
Sincerely,
DesertAdmin
Any input/insight on anyonelese having this issue would really help.
Am I the only one with a 10Gbps as an ISP feed?
We are just pushing the limits of 1Gbps and it seems like our feed can not due more than 1Gbps due to using Mikrotiks. I think this is a real issue. Perhaps it is the ROS 6.41.2 or the firm of the 6.41.2.
Again some helpful insight would help.
-Sincerely,
DesertGate Internet
#1stWorldProblemsIsA10GbFeed
There are only one way to test 10Gbps connection in house:
https://wiki.mikrotik.com/wiki/Manual:Performance_Testing_with_Traffic_Generator
any external single connection tests like speedtest.com etc, because of round trip time will be limited to sub 1Gbps speeds anyway.
There are only two ways to get such speeds with single connection - have veeeery small RTT - so basically next box in same building, or Jumbo frames (whole network need to be configured for MTU 9000 or above)
with real client traffic - ~400k entries in conntrack table i have ip to 8Gbps on 10G interfaces on NAT/queue router, and we use DAC cables from switches, so as far as i can see , your testing method is not suitable for 10G,
Thanks for the reply that is an interesting link test measuring way. I do not want to stress the boards I have approx 2000 real connections going through and we do not see greater than 1Gbps on an all 10GB SFP+ network from end to end.. So I have just recently started again to run the Bandwidth test from one RB1072 to the next that is connected with DACs. Nothing between.
I am not testing with any outside speedtest site. I understand their limitations. I am running the bandwidth test from one 1072 to the next on the same rack. I expect to see at least 2Gbps. but with DACs SFP+ I am now only seeing 1Gbps at best I think this is due to a firmware update some time recently. I use to be able to test higher now it is no longer reporting greater speeds.
Perhaps someone knows which version is best to run before it was broken. That may help me better. Thank you
-Sincerely,
DesertAdmin
Bandwidth test is limited to single CPU core, it can’t generate more than ~2Gbps and even then with many connections.
It doesn’t test more than 1Gb at 4:30am with only 300Mbps that is pushing through. It seems to top off at 1Gbps. But I really saw (not less than 6 months ago) the approx of 2Gbps when doing a speedtest (Bandwidth Test BTEST tool from MT to MT ) on my older RB1036s that had the 10Gbps SFP+.
Seems like it was a ROS or firmware change.
-Sincerely,
DesertAdmin
Ok I am confused where to change the MTU size limit. I am RTFM but I am getting confused any place to point me to would help. on how you changed the MTU size on a bridge it seems to be gray’ed out.
-Sincerely,
DesertAdmin