Hi, I wanted to check, the following
The company is a small ISP, with 400 users, all wireless (nanostation).
The main router for bandwidth control is an RB1100AHx2, with dhcp and simple queues. All subscribers connect to the internet through NAT.
The connection speeds are 1M / 3M and 1M / 5M (up / down).
The query is what queue size they recommend to use,
default uses the default-small queue type (pfifo)
name = “default-small” kind = pfifo pfifo-limit = 10 -
If I increase that value, do I improve navigation? because I see on the internet that some alter those values.
I know that the larger the size of the tail, the less drops, but the greater the latency.
What values would be optimal? I would like to know the experiences of some isp-
Thank you
To calculate the value, multiply connection speed by maximum latency. For example, if you decide the maximum latency you want to add is 100 ms, and your average connection speed is 4 Mb/s:
queue length = 4 Mb/s * 100 ms
queue length = 4 Mb/s * 0.1 s
queue length = 4 Mb * 0.1
queue length = 0.4 Mb
queue length = 50 kB (converted bits to bytes)
Now pfifo takes its value in packets instead of bytes. Internet packets range in size from 64 to 1500 bytes. Typical home usage is large packets, especially when the connection is saturated, which is the only time the queue gets involved. Let’s assume 1000 bytes average. Then that 50 kB works out to be 50 packets.
For your case, I recommend using the red queue type instead of pfifo. It adds some complexity to try and be gentler to traffic when the customer is maxing out their connection. Is it still very simple compared to modern methods like CoDel but it’s better than FIFO.
thank you , for your help , but I can’t understand how you get to the result -
What part don’t you understand?
queue length = 0.4 Mb ??
queue length = 50 kB (converted bits to bytes)
??
@wmackay ,
it is very difficult to implement queues of the red type. My environment is a small isp with 2 plans
1MB / 3M (up-down) and 2M / 5M (up Down)
thanks