Router reviews - different packet sizes

[System Error Message]

Looking at everyday traffic, especially web traffic which i have so many of i noticed that most of the traffic dont use the full MTUs in internet from the device level. While normal downloading and streaming will take advantage of this i am seeing the bulk of my traffic on my WAN interface on average about 1000 bytes per packet on download. My upload traffic on average bytes per packet is ludicrously low in comparison to download.

I would like to suggest LAN-WAN and vice versa tests using 512 and 1024 byte packets a bit like what mikrotik does with their performance testing but mikrotik doesnt do NAT performance testing. True router performance is usually measured with packet rates as this helps to actually measure how much overhead when using vlans and PPPOE with internet connections. As some experts here seem quite surprised that many ISPs that provide gigabit/multi gigabit or fiber optic internet still use PPPOE.

Using PPPOE and VLANs does drop the MTU so a packet rate test would be more accurate as a speed measure. PPPOE+VLAN+NAT testing is actually quite easy to set up and perform as long as the test equipment is capable of wirespeed PPPOE (perhaps a high frequency dual/quad core router/server)?

For example if using 1500 byte packets you need less than 7000 packets to fill 1Gb/s but you will need 2 million 512 byte packets to fill 1Gb/s which gaming and services (like NTP, DNS, commands) tend to use.

 

[thiggins]

Valid point not only for routing but also WLAN throughput tests. But can't afford to increase test time even more than it already is.

Why don't you run some tests with iperf/jperf and post the results?

 

[System Error Message]

Valid point not only for routing but also WLAN throughput tests. But can't afford to increase test time even more than it already is.

Why don't you run some tests with iperf/jperf and post the results?

Im not sure the networking gear i use is the kind people who come here are looking for. For example how many here would actually buy a CCR1036? I dont even have enough x86 hardware to test the CCR1036 speeds as well.

I dont have a faraday cage for optimum wifi testing but i could try WAN-LAN for the asus routers i have.

Correction: I forgot to divide by 8.

 

[L&LD]

The MTU must not be confused with the minimum datagram size that all hosts must be prepared to accept, which has a value of 576 bytes for IPv4[2] and of 1280 bytes for IPv6.[3

https://en.wikipedia.org/wiki/Maximum_transmission_unit


SEM, isn't that then contradicting what it says in the quote and link above?

 

[System Error Message]

https://en.wikipedia.org/wiki/Maximum_transmission_unit


SEM, isn't that then contradicting what it says in the quote and link above?

No, what i am talking about in this thread is the actual performance of a router which is to do with how many packets it can process in a second. This isnt to be confused with MTUs because MTUs only mean the maximum whereas every packet usually varies in size based on the payload.

Problem is people dont really understand router speeds in packets per second, they only understand the speed of routers in the standard 1500 byte packets when the variances of WAN mediums also mean varying MTUs , especially if PPPOE is used as the MTU will be below 1500. However most traffic does not use the reach or exceed the MTU of WAN even PPPOE so if you test using minimum packet size to get performance in packet rate this would be more accurate in finding out the minimum and maximum speeds the router can achieve.

There is a maximum speed the router can achieve not in packets per second but when a packet size is too big that it fills up the cpu/memory bus as you can actually do in a mikrotik CCR by using 65 KB Packets .

 

[L&LD]

Problem is people dont really understand router speeds in packets per second, they only understand the speed of routers in the standard 1500 byte packets when the variances of WAN mediums also mean varying MTUs , especially if PPPOE is used as the MTU will be below 1500. However most traffic does not use the reach or exceed the MTU of WAN even PPPOE so if you test using minimum packet size to get performance in packet rate this would be more accurate in finding out the minimum and maximum speeds the router can achieve.

There is a reason it is a standard (as are the lower than 1500 byte standards when PPPOE is used too).

This information you see would not be relevant or useful to me or my customers. Except in a theoretical way.

When the MTU is set at the default and tested at that setting for all routers, that is a useful metric for all.

I know you want to try to 'break' your network equipment to know it's limits. But that information is very specific and not that applicable, to most.

 

[System Error Message]

the MTU is only the maximum per packet/frame, most packets are smaller than the standard MTUs so testing using minimum sizes would show a more accurate performance as packet rates as that is the performance of a router, the rate of packets it can process, MTUs dont really matter in this regard as long as the busses arent the bottlenecks.

 

[L&LD]

the MTU is only the maximum per packet/frame, most packets are smaller than the standard MTUs so testing using minimum sizes would show a more accurate performance as packet rates as that is the performance of a router, the rate of packets it can process, MTUs dont really matter in this regard as long as the busses arent the bottlenecks.

I understand that the MTU is just a 'suggested' value. But what you're suggesting would make comparisons of different routers more complex, not less.

As tested now, they are comparable (at a single data point). Different routers being better at different packet sizes won't help me choose one router over another, there are more important things to consider, imo.

 

[System Error Message]

I understand that the MTU is just a 'suggested' value. But what you're suggesting would make comparisons of different routers more complex, not less.

As tested now, they are comparable (at a single data point). Different routers being better at different packet sizes won't help me choose one router over another, there are more important things to consider, imo.

This isnt about testing different sizes, this is to test using the minimum size only to find the absolute performance of the router.

 

[sfx2000]

I dont have a faraday cage for optimum wifi testing but i could try WAN-LAN for the asus routers i have.

I've got a 40 acre location still under lease in northwestern North Dakota (was the site of my mesh networking startup) - It's a totally clean zone for 2.4 and 5 GHz - stop by the house, and I'll give you the keys to the gate...

It's got a 1GB WAN (licensed RF/Microwave link) connection with battery backup (we paid a lot for that one, and it's still hot, 4 years later, just checked). Good power out there if needed - and you'll probably still find my nodes out there, cold-iron (turned off) - ATT and Verizon have LTE coverage - but if you want/need a free space test area to explore different options, this is the place

I'll need to know in advance - I'm bouncing the microwave link to a couple of nearby farmsteads - just need to let them know...

 

[System Error Message]

we're quite apart separated by lots of sea

 

[sfx2000]

we're quite apart separated by lots of sea

Offer stands for you...

 

[System Error Message]

lol thanks, you're quite lucky to have 1GB WAN, here in the UK internet speeds arent fast enough to register any CPU usage on my CCR1036

 

[RMerlin]

lol thanks, you're quite lucky to have 1GB WAN, here in the UK internet speeds arent fast enough to register any CPU usage on my CCR1036

All I have access to here are CR2032s

 

[System Error Message]

All I have access to here are CR2032s

you mean the coin cell batteries :O?

@RMerlin do you have access to chemicals? You could create your own 12V car battery using chemicals as a power backup but you will need the casing and ventilation. There may not be much sun in canada but solar is also an option and ubiquiti has been actively promoting their solar products in canada.

 

[L&LD]

I know you want to try to 'break' your network equipment to know it's limits. But that information is very specific and not that applicable, to most.

This isnt about testing different sizes, this is to test using the minimum size only to find the absolute performance of the router.

My post #6, quoted on top, shows I understand what you're asking for.

You're not getting that my reply has been some form of the last sentence in the top quote for a few posts now.

 

[sfx2000]

lol thanks, you're quite lucky to have 1GB WAN, here in the UK internet speeds arent fast enough to register any CPU usage on my CCR1036

It's not a residential type of link - it's a dedicated microwave link that I negotiated about 5 years ago, with SLA's and what not - latency is a bit higher than I would like, but it's basically the same type of link one would see on a cellular network tower... I've got a couple of months left on contract, and in negotiations with the two farms that depend on that link (I've got 900MHz point to point links to them, if I fall out, they lose connectivity) - right now, they pay me, and I pay for the link...

 

[sfx2000]

It's not a residential type of link - it's a dedicated microwave link that I negotiated about 5 years ago, with SLA's and what not

And while it would be nice to have this link at home - sadly not the case, as it is 1200 miles away... at home, I've got residential cable at 150/10, which is probably good enough for most purposes...

The costs to build and run a link like I have at the "farm" is insane for any home user, and for many small businesses in a metro area - but once I broke the link out to the other two locations, and come up with a reasonable pricing policy - nice thing, back in the day, no caps on bandwidth used... and my two customers are pretty satisfied with it.

 

[System Error Message]

And while it would be nice to have this link at home - sadly not the case, as it is 1200 miles away... at home, I've got residential cable at 150/10, which is probably good enough for most purposes...

The costs to build and run a link like I have at the "farm" is insane for any home user, and for many small businesses in a metro area - but once I broke the link out to the other two locations, and come up with a reasonable pricing policy - nice thing, back in the day, no caps on bandwidth used... and my two customers are pretty satisfied with it.

perhaps you should live on the farm instead . At least its not comcast or AT&T lol.

Since you have a farm you could probably use solar to run many things. The area you would have you could do quite a lot.

 

[sfx2000]

Since you have a farm you could probably use solar to run many things. The area you would have you could do quite a lot.

We do for the community - but I like living in San Diego