I’m using Rb750Gr3 and SwOS (CSS326-24G-2S+RM) to handle 16 LAN users and 4 wireless AP (around 30 WLAN users) in 3 floors building.
Since I changed to this infrastructure, the purpose is to keep internet usage quality (low latency and fixed download/upload speed.
But it always have high awful quality (hight latency) in the crowded hours, I couldn’t even find out the root cause for such condition by what reason or which certain user.
So I’m planning to upgrade my router from Rb750Gr3 to x86 system.
I knew there’s lot suggestion which MikroTik had done lots optimization no matter on CPU/Memory/ ..etc which is x86 system don’t have.
But my standpoint here is my X86 system spec should be able to handle communication better and have great improvement on internet access quality.
My x86 system spec is listed as below.
CPU: Intel® Atom platform X5-3940 1.6GHz (4C4T, TDP 9.5W)
Memory: DDR3L 4GB x1
Storage: mSATA SSD 128G x1
LAN: Giga LAN x2 (Intel i210 chip)
Here, my first question would be how to build up my environment.
Option 1: Install RouterOS software version directly as master.
Option 2: Install Win10 OS, but use VMware to install RouterOS software to manipulate (Win10 OS could be used as wireless AP master).
But I’m still wondering whether option 2 would have the same performance with option 1 or not.
Or even both option 1 and 2 won’t be better than Rb750Gr3.
Is anyone could give me some advices/suggestion?
Or there’s way to improve latency issue on Rb750Gr3? By script, specific setting by limiting P2P or…others.
Really appreciate your kindly feedback.
Thanks.
P.S. Users would use every kind of scenario, including online gaming, video streaming, internet surfing, …etc. All kind of things you could image.
Based on your description, it’s not clear where the problem is. Are you routing what, megabit, two, ten, hundered, gigabit? If you have almost fifty users, they can easily max out even a fast connection, and then of course latency goes to hell. Do you have connection with guaranteed speed from ISP? If so, you might limit it yourself a little below ISP’s limit, play with queues and priorities and it could work reasonably well.
It sounds like more of a configuration problem vs a hardware problem. If you cant determine the cause of latency now then changing hardware wont be accomplishing anything.
tool profile allows you to see the routers resource usage to determine if its being overloaded in some way. Too many firewall rules, Open DNS resolver etc.
Are people torrenting? Causing a massive number of connections? Are some users consuming all available bandwidth? Saturation is generally the most common cause of high latency in a LAN.
Try queuing the users and see if it improves. If you queue you may need to modify or disable fast track in the firewall.
I use single 100M/40M ISP Host w/o PPPoE, then plug to RB750Gr3 and DVR camera directly.
Then a SwOS hub (1Gbps) linked behind my ROS. https://imgur.com/r0nSORP
Based on my understanding, I haven’t reach the maximum of my bandwidth.
It always be around 10-25 MB/s as below graph. One of the reason might be there’s not that many user are using it due to terrible quality. https://imgur.com/KkczK1y
The random IP is controlled by Rb750Gr3 via DHCP.
All the other AP use LAN to connect for spreading WIFI signal only.
About ROS setting, I’m not really familiar with the settings.
But I had disable “fasttrack connection” in Firewall Rule and done with 8M/1M limitation for each user. https://imgur.com/oVYa6hX
No Script or Layer7 Protocols had been set.
I don’t really find where I set the connection limitation, but it should be 100 connections for each of them.
As my understanding, the P2P will some how constrain by less peers.
In the end, I had started to use EMCO Ping Monitor (free software) on x86 system to capture the latency log since yesterday.
Here is what I have for now. https://imgur.com/iz9gPLj
Anything would help you to know more about my environment just be free to let me know.
Thanks again.
Check CPU usage in System->Resources or Tools->Profile
Do you know what exactly 100M/40M means, i.e. if it’s guaranteed that you always have the full speed available to you, or it’s that these numbers are maximum, something you should usually get, but in reality a connection could be shared by several customers, and not be fast enough to give maximum speed to all of them at the same time?
Most of the time, the CPU usage is under 10%.
I use “netwatch” to monitor latency with Google DNS ( 8.8.8.8 ), the last few times I found the latency time over 20ms is below (once a minute):
a) [2018/6/13 17:23] CPU loading is 29%, and back to under 20ms when CPU loading 11% in next ping value. ← Probably caused by bandwidth (high CPU loading)
b) [2018/6/12 22:51] CPU loading is 10%, and back to under 20ms when CPU loading 7% in next ping value.
c) [2018/6/12 07:25] CPU loading is 5%, and back to under 20ms when CPU loading 8% in next ping value.
d) [2018/6/12 17:50] CPU loading is 35%, and back to under 20ms when CPU loading 28% in next ping value. ← Probably caused by bandwidth (high CPU loading)
=>Item d is the most special case that I ever had, becuase my traffic graph had monitored the download speed reach the peak around 80mb/s.
It’s the highest CPU loading since I used “netwatch” in these 2 weeks.
For 100M/40M, I think it should be the full speed which I can get. I usually could get this download/upload data on speedtest without bandwidth limiation.
But of course it’ll differ by current user quantity or behavior.
Does it mean I might be able to fix this issue if I could enhance my bandwidth from ISP?