I’ve had my RB2011UiAS-2HnD-IN for a few years now and have had no complaints until now. I moved to my new house and upgraded my internet speed from 150/10 to 400/20.
I ran fasttrack off and on at this speed for couple months and saw there were some updates to RouterOS and that I had been a little slacking on upgrading my firmware (as in firmware had never been upgraded since i had it…whoops!).
I update the RouterOS and the Firmware, including rebooting for both my 2011 and WAP AC and all seems to work fine at first glance.
Then I noticed that even with Fasttrack enabled, I was only getting between 150-200mbps max download with the exact same config as before with only about 5% CPU at idle, 30~% under load with fasttrack enabled. I even install a previous backup from a time when I know i was getting full speed just to see if I messed something up, but the speed issue remained. When I connect directly to my modem from my computer, I get about 440/22, so I know the issue is not on my ISP’s side. I even tried to move all other wired connections to the other switch (the 100mbps ports) to rule out any interference from other ports on the same switch.
The internet speed also seems to be inconsistent when running through the 2011 through fasttrack: it sometimes starts out around 250-280mbps, then slows to 150-200 when using speedtest.net. Sometimes it starts at 150mbps and goes down to about 110mbps. With fasttrack disabled, i get about 73mbps down with 80% CPU usage.
I’m running 6.43 currently on routerOS and firmware
I have about 8 wired devices and about 10 wireless devices with about 25 Mangle rules, 3 nat rules (2 for plex server, 1 for masquerade), and a total of 17 firewall filter rules with fasttrack as the first after an invalid reject rule.
I’ve tried fasttrack set to my machine’s marked packets, marked connections, and then all established and related connection states to try and bypass all filters and rules.
I’ve disconnected all other devices and cables, disabled WLAN, reset configuration…No Dice.
I’m not at all a mikrotik expert, but I’d appreciate any help!
I’ve read that multiple times, but it’s not clicking for me…isn’t hardware offloading supposed to improve performance? So should i reformat and create the setup without master/slave setup from the beginning? I’ll try and re-read it a bunch to see if I can figure it out.
You should verify that hardware offloading is being enabled (indicated by H indicator on bridge port) and verify which functions are supported with it for your device’s switch chip and confirm you’re not running with any of incompatible options enabled.
In the bridge ports, the wan port was not in the bridge (i assume this is correct, since i can’t add it to the bridge).
I tried disabling the hardware offload that my computer is connected to and still get the same result. I disabled Unicast Flood, Multicast Flood, Broadcast Flood and still the same. Now, the speed test starts at my max download (400mbps) for a brief second and then quickly ramps down to 200mbps for the rest of the test. (attached
CPU maxed at ~40% during 400mbps internet speed test download, then dropped back down to ~25% for the rest of the 200mbps average. When transferring files on local LAN at gigabit speed, CPU goes up to ~65% for 950mbps speed going from a hardware offloaded port to non hardware offloaded. On hardware offloaded port to hardware offloaded port on the same gigabit switch series, cpu doesn’t go above 6% for same gigabit speed transfer, so hardware itself seems to be working fine.
Any idea on what settings might be causing the drop in speed after the initial good speed?
Round-trip latency with too small TCP window? Not that you can do much about TCP window size for random application …
Any network device on the way between connection peers will add some delay. It sometimes really surprises me how can seemingly neglectable increase in delay (a few 100 microseconds) drastically change TCP performance.
I’ve got cable internet connection. Setup is internet<->cable modem<->rb2011<->laptop with cat6 cable all ethernet connections at 1gbps full duplex. Ping times to bandwidth test server are under 20ms.
I took ethernet coupler and hooked my laptop ethernet cable to the ethernet cable from the modem to bypass the router but still using same cables and I get full consistent 400mbps down going straight through modem, so wiring is ok. I also tried to clone Mac address of my laptop to router after connecting straight to modem.
Not sure what setting I’d need to check? MTU is 1500, 1598 max l2 MTU (thought packets might be getting fragmented). Guess I can change wan port and see if problem persists?
So I did some messing around:
Hops to bandwidth server is 8, total RTT is 12ms to bandwidth server.
So I started changing things:
I disconnected all other wires/wifi. Same speed
I reset RB2011 to vanilla WISP AP/Home AP configuration with only modem and laptop connected via ethernet. Same speed.
I used vanilla configuration to change ethernet port of WAN to Eth2 instead of 1. Same speed.
I’m thinking it had something to do with updating my firmware to current, since speed was good before that time. Is it wise to leave RouterOS updated but firmware not updated?
I got a response from mikrotik talking about how queues impact TCP single connection performance and i should just be able to get normal real-world speed outside of the synthetic speedtest.net test. I tried to download a large file, such as a ISO file via HTTP or torrent and still was stuck at about 200mbps maximum. I also tried to download a game via steam and still wasn’t able to get above about 200mbps.
I went to dslreports.com and performed a speedtest and got a full 450mbps or so speed test .
I understand how queues can impact tcp single connection performance, but I had fasttrack enabled and got the same results. I also reset the router to vanilla WISP AP and also Home AP and only had laptop and modem connected to the router with no queues or mangle rules and still only got 200mbps. Any ideas on how to improve TCP single connection performance?
I hate to look at a CCR or RB3011 or something more expensive to get the performance that i was already getting on the rb2011 until some strange gremlin reduced my speed.
What firmware are you running? Could you share your configuration? I used to get great wan speed but still get great lan routing speed. I tried resetting to stock wisp ap defaults as well as home ap default config and can’t get any better on that speedtest.net test, so I thought I eliminated bad config.
Sorry, just looked at the speedtest results again, it was 812Mb down and 97Mb up as the link was a 1Gb/100Mb fibre link. About a month ago, I have reset the 2011 to Factory Default and downgraded the device to my Home Lab environment, using a HAP AC2 as “production” unit at home now.
If you don’t mind placing a copy of “export hide-sensitive” of your config here, we can have a look and make suggestions
These are both screenshots for speedtests i get with this configuration. CPU doesn’t exceed 45% with the faster result speedtest. Slower speedtest cpu doesn’t exceed 26%.
Do I understand it right and correct me if I’m wrong, but after resetting to vanilla WISP AP, the download speed is still lousy?
Please check the Routerboard “Current Firmware” is 6.43 and/or re-install the 6.43 packages.
Yes router os and firmware are both updated. I posted some screenshots earlier today with dslreports speed test showing normal speed (multiple connections?) but any other speed test (single tcp connection?) shows slower speed. My modem is not a router, so no bridge mode possible there. I’ll try and reinstall firmware and router os.
Firstly, when pasting config, please place this between the code brackets, makes it easier to read and not such long posts.
Then, I am a bit confused, you are showing 2 speedtests screens, one with 158Mb download and another with 448Mb download. the 448Mb is not to bad.
In current config, you have many Q’s and Mangle rules, at most I had about 3 simple Q’s so difficult to compare, but you also have “passthrough=yes” on almost all Mangle Rules, that is not correct and will eat CPU cycles which will affect routing speeds.Have a search through @sindy’s posts, some excellent explanations.
I will also reorganize the firewall filter rules, what I had in ordered list (Very basic but very effective I think):
chain forward action drop connection invalid
Chain forward action fasttrack est, rel & untracked
chain forward action accept est, related & untracked
chain forward action accept connections new from LAN
chain forward action accept connection state dstnat from WAN
chain forward action drop all from WAN
chain input action drop connection invalid
chain input action accept est, related & untracked
chain input action accept connections new from LAN
chain input action drop all from WAN
Thanks CZfan, I fixed the code view. CPU never exceeds 45% or so when running at that 450mbps speed and my connection is going through fasttrack. Both of those screenshots were different bandwidth tests with that same configuration that I’ve been running for 2 years. One of the connection tests runs through multiple TCP connections is what mikrotik support was telling me, the other is only a single TCP stream. I used to get these speeds normally from speedtest.com when i first moved to this house in june, using that same configuration with fasttrack, but now i’m not. The only thing that changed is i updated my firmware and routerOS to current.
.
