Hello, we have problem with 10G speed when to connect CRS317-1G-16S+RM by SFP+ direct attach cable (S+DA0001 or S+DA0003). We have ~ 1.34 Gbits/sec. We have tried swith off “Auto Neogation” but it did not help. Server1 to Server2 connecting by SFP+ cable gives true speed 10.9 GBytes. Connection diagram in attached. Please help, thank!
[attachment=0]mkt-crs317.png[/attachment]
Is it true that two CRS317-1G-16S+RM devices are involved in this test?
Why not testing on a single device first?
If really two are involved, then they better should be in their own LAN (ie. IP should be something like 192.168.88.1/24 and the other should be 192.168.89.1/24). At least for the testing.
It seems the traffic goes over your 1G gateway address 
Hmm. maybe not as it’s 1.34Gbps.
I think it’s a routing problem.
The outputs of these would be needed for further analysis:
/ip route print
/ip address print
And the IP+mask of all involved devices / interfaces.
Another observation: it seems you used iperf for testing, but does it not have also a SUM line under the ID heading?
Thank you!
Is it true that two CRS317-1G-16S+RM devices are involved in this test?
yes
Why not testing on a single device first?
/ip route print
/ip address print
[admin@MKT-B.1 (Primary 192.168.88.1)] /interface bridge> port print
Flags: X - disabled, I - inactive, D - dynamic, H - hw-offload
0 H ;;; defconf
sfp-sfpplus12 bridge yes 1 0x80 10 10 none
[admin@MKT-B.2 (Secondary 192.168.88.2)] > /interface bridge port print
Flags: X - disabled, I - inactive, D - dynamic, H - hw-offload
1 H ;;; defconf
sfp-sfpplus12 bridge yes 1 0x80 10 10 none
@nickkk, I can just suggest this: use iperf on PCs for performance tests, not the integrated traffic generators on the routers or switches
as this creates additional CPU load which then is missing for the device itself to perform its routing/switching job.
And: do the test first w/o VLAN, and on a single device with 2 PCs attached to it.
Maybe someone else can help further as I have to give up b/c VLAN is really not my field.
(VLAN for me is like old BASIC spaghetti programming with many GOTOs Sorry, couldn’t resist! )
Btw, I too did such 10G performance tests on the CRS326 switch between two servers attached to this switch via 10G SFP+: the performance was about 9.5 Gbps, ie. ok.
Test on first image by iperf.
Thank you.
Hello, I installed two new CRS-328 without any vlans and firewall rules. They connected by “SFP+ direct attach cable (S+DA0001)” too.The result is the same ~ 800 Mbps. I can test only through “Mikrotik Btest tool” for it now. But I think iperf will give the same result as it was in the first test with servers.
I want to try changing the cable to optical, but now they are not. I would still like to deal with “SFP+ direct attach cable” , we have ~ 20 connects.
I was facing similar throughput issue - I had wrong bridge config. Make sure u use only one bridge and Hardware offload vlan filtering on on bridge.
This I get about 9,8 gbit using xen motion or iperf test (server needs power)
Hello, thank for help.
There is only one default bridge1 on each mkt
[admin@MKT-G.1 (CAM Primary)] > /interface bridge print
Flags: X - disabled, R - running
0 R name=“bridge1” mtu=auto actual-mtu=1500 l2mtu=1592 arp=enabled arp-timeout=auto mac-address=:C2 protocol-mode=rstp fast-forward=yes igmp-snooping=no auto-mac=yes ageing-time=5m
priority=0x8000 max-message-age=20s forward-delay=15s transmit-hold-count=6 vlan-filtering=no dhcp-snooping=no
[admin@MKT-G.2 (CAM)] > interface bridge print
Flags: X - disabled, R - running
0 R name=“bridge1” mtu=auto actual-mtu=1500 l2mtu=1592 arp=enabled arp-timeout=auto mac-address=:BA protocol-mode=rstp fast-forward=yes igmp-snooping=no auto-mac=yes ageing-time=5m
priority=0x8000 max-message-age=20s forward-delay=15s transmit-hold-count=6 vlan-filtering=no dhcp-snooping=no
Tried vlan-filtering=no to change to vlan-filtering=on , bandwidth was less
In your first schema red square should become a router! Transfer routing/l2tp/filtering to him. Your switch is L2 device(access level). And your bridge config incorrect.
https://wiki.mikrotik.com/wiki/Manual:Interface/Bridge#VLAN_Example_.231_.28Trunk_and_Access_Ports.29
Red square in first schema remained from the drawing program “Xmind”. This connection does not exist. Sorry if I misled you. All subsequent tests are conducted between two mikrotiks with default configuration.
Corrected first scheme
