Hello guys,
I read a lot posts before writing this, but I didn’t solve my problem.
I bought 2 pairs of Netmetal5 (2 with RB922UAGS-5HPacT-NM and 2 with RB921UAGS-5HPacT-NM). I upgraded the RouterOS to version 6.37.3. I would like to create a stable ptp link working as Ethernet bridges (eth<->wlan<->wlan<->eth). No problem of signal (there is LOS and distance is 500m). I configured the system in WDS (bridge+ station wds) and nstream (nv2 has worst performance) in 5GHzAC-only band, with channel size 80MHz Ceee (I have the wireless channel established with a capacity of 1300Mbps). I’m using a profession device to test the behavior of the link(i.e., a router tester).
I would like to replicate the performance reported in the specification:
Bridging none (fast path): 986.1Mbps (with packets of 1518 byte), 962.2Mbps (with packets of 512 byte) and 244.6Mbps (with packets of 64 bytes)
During my tests I’m sending packets with 1500 bytes, but I was unable to exceed 120Mbps on the link. there is some problem in my configuration?
The CPU load is in the range 37-60%. I tried to disable all the other packages not required. Moreover sometime I have a burst of frame loss (minimal but present) also if the link is stable.
Do you have some suggestion?
Please let me know if you need further details.
Thanks a lot for the support.
Looking at the first images you posted your ccq is a little low. you possible have some interference or you need to turn down your power on your TX. have you tried different wireless protocols.
Ethernet test results
...
1. All tests are done with Xena Networks specialized test equipment (XenaBay),and done according to RFC2544 (Xena2544)
...
4. Test results show device maximum performance, and are reached using mentioned hardware and software configuration, different configurations most likely will result in lower results
These tests use the ETHERNET interfaces, and they measure aggregated speed to assess the core performance.
In the case of your 922, this means traffic from ethernet to SFP, not via wireless only.
@dgnevans: forget about the first link (it was under installation with no finalized setup-up), just consider the second one that was up and running.
@hengst: the max throughput without vlans that I’m able to reach is around 120Mbps of Ethernet frames (no TCP, not UDP with link capacity of 1.170/1.300Gbps). With TCP I’m not able to exceed 50Mbps. It seems that the processing of the packets is too CPU hungry… Why using a VLAN should solve my problem?
@docmarius: thank you for highlighting the magic word (“ETHERNET”) in the specification: I assumed that this Etherent mean eth-to-eth tests were between 2 different Netmetal5 devices. If I well understood, the specification test has been conducted involving the ether1 and the sfp1 interfaces. But in this case why I should use a Netmetal5 as Ethernet switch? It is a pity to use only the 10% of the link capacity…
I believe there might be something wrong with packets from wireless interface “through” the ap device. We are testing a lot trying to reproduce a clear view on that. I was curious if vlan would be differrnt then all going thru bridge. Your wireless id clearly showing its connection rate going up to 1Gb. so your max tcp trhoughput through the ap should be capable of lets say 450Mbits one way (half duplex), so where is it?
Between A and B it sends 350Mbps TCP in BTEST (cable)
Between B and C it sends 170Mbps TCP in BTEST (wireless)
Between C and D it sends 350Mbps TCP in BTEST (cable)
Between A and D sends only 70Mbps in BTEST (cable + wireless)
I would not use nv2 here, as thats more useful to ptmp. I think nstream will be a better choice. Also check ack settings with your link calculations if you use ack.
