Short version:
I am no longer able to get the expected performance out of the RB5009 that I used to. ISP provides 1.2Gbps (overprovisioned to 1.4Gbps) from cable modem to 2.5GbE port on RB5009. Can only get a maximum download of ~750Mbps from any of the remaining GbE ports OR from the SFP+. (From the 1 GbE port, should at least get 950 Mbps; on the SFP+ using a 10GBase-T module to connect to different 10GbE ports and clients connected through it, used to get 1.4Gbps)
This is on a 1.5 yr old RB5009 (PoE version), as well as a brand new RB5009 (non-PoE) with default config of ROS 5.8 out of the box (or upgraded to the latest ROS).
Any ideas on where to start?
Long version:
I purchased my first RB5009 back in September 2022 when it first was available. It replaced my RB4011 router-on-a-stick setup with a more standard one with the 2.5GbE port connected to my cable modem, and the SFP+ used to feed the rest of my network.
At the time, it worked exactly as expected, and I was able to fully utilize the Comcast/Xfinity 1200/40, and later 1200/200 speeds.
At some point, my first RB5009 was bricked after a RouterOS update, and I ended up replacing it with a 2nd RB5009 (the PoE version, because it was available where the non-PoE was not). I basically recreated the same configuration and had expected performance.
Over time, my config was expanded to support multi-WAN, ZeroTier, and additional services.
Starting earlier this year, I noticed I was no longer able to hit the 1200/200 (really 1400/240 with overprovisioning), but assumed it was something on the Xfinity side of things. They recently performed some “network maintenance” in the area and afterwards, the performance is still not what it used to be, with ~750Mbps maximum download.
I did some initial testing with no improvement. Then, I order my 3rd RB5009 (non-PoE) and right out of the box (RouterOS 7. it still seems to be limited to the same ~750Mbps maximum download.
I’m not sure where to go from here, since the RB5009 used to be able to hit the expected throughput before.
If it used to work before and not anymore now, the most logical reason would be … changes you made to config over time.
My reasoning:
If it would be the device itself, this place would be swamped with reports stating the same issue. It is not. (I have 2 myself, zero problems)
If it would be SW version, same remark. Though it never hurts to upgrade to latest stable in most cases.
If it would be a HW problem, I assume chances are higher the output of the device would be zero, not lower performance.
So …
Time to put your config on the table.
Terminal
/export file=anynameyouwish
Move file to your PC, remove any remaining sensitive info (serial, passwds, public keys, public IP, …)
Copy-paste file in between <__code> quotes for easier readability.
The only true testing is disconnect the ISP and put some testing-PC and perform iPerf trough the RB5009 or some other test-tool.
I would not rule-out your ISP…“cable modem” is a shared medium and you never know exactly what happened. You also probably do not have performance guarantees on your current Internet subscription.
Do you really think your ISP has enough bandwidth enough from the street-cabinet terminating a bunch of customers if each of them is pumping “the limit” with their speed tests ?
You are not running any “multi-wan” stuff right ? Plain & simple 1 ISP now, now QoS-rules etc?
This router is not capable of 2.5Gbit/s. Not in real life condition
With fasttrack and all optimizations tested it with iperf, PPPoE speed max 1.6Gbit/s because CPU Mhz/software limits speed.
Dearr GolemPL, ObliteRon has a 1.2gbit ISP connection and until some point was able to fully utilize the given bandwidth. No need to comment here, that this RB is not capable of 2.5gbit/s. It adds nothing to the issue resolution.
@GolemPL, it sounds like you might have some heavy firewall rules, queues or software encrypted tunnels that are causing all the traffic to be CPU-bound.
Testing internal BTEST between RB5009 and AX3, I can completely fill that 2.5Gb line between both devices. Plenty of power to spare on both ends.
Internal test towards loopback port, it gets to almost 10Gb (9.6, if I recall correctly)…
Cable-modems do not use PPPoE in general. Plain DHCP across the ethernet and that’s it.
So that is already a hassle / limitation / overhead less to deal with.
I came back to this after leaving it alone for a few weeks.
On the new RB5009, I’m not quite sure what I did different but I was able to get expected throughput out of it.
On my existing RB5009, the issue was that I had tried to address bufferbloat at some point defining queues. When it wasn’t working as expected, I had disabled the simple queue I created in Queues > Simple Queues, but didn’t disable the queues created in Queues > Queue Tree. Once those were disabled, throughput is as expected.
The main thing that tipped me off is that in Queues > Interface Queues, the Active Queue Type for ether1 (my primary WAN) was showing queue-tree, instead of the assigned queue type.
it still seems to be limited to the same ~750Mbps maximum download.