Question likely not in a 'good' or 'proper' part of the forum so please move if needed!!!
At what point would a network be considered saturated?
I'm working on a project where a lot of information would get shuttled around on my internal LAN.
Wondering how much headroom a 100 Mbit level of network gives.
(Sorry - - - I tend to plan in large scope whilst I am working on what I need to do today or pdq so I ask questions about things that may not be at all up close. Also really couldn't find much info online. Hopefully not a frivolous, or considered so, question. )
Expect around 94Mbps on a 100Mbps link. But it always depends on the packet size. With bigger packets you can achieve higher throughput.
@trendy is correct, but I'm wondering if there is more to the question in the OP?
For example, a network can become internally saturated if there is a significant amount of broadcast traffic. Typically that is caused by too many devices on a single network, but the definition of "too many" depends in large part on the types of devices, services, and data that is flowing.
Another important element is that on a wired switched network (where all devices are connected on the same switch), two devices can talk to each other on the internal network without impacting the bandwidth of 3rd device connecting to the internet and/or to another internal device. This becomes more nuanced when multiple switches are connected together, and it is obviously not necessarily true for wireless because the devices need to share the airwaves, of course.
All of that said, @trendy's answer is spot on and applies to any individual link running at 100Mbps, and certainly if the discussion is related to saturating a 100Mbps connection to the internet, that constitutes an individual link.
My experience with my old 100Mbit devices I had was that they where actually never built for 100Mbit usage?
This speed was never really a industry standard as Gbit network was so a lot of home devices got a 100Mbit RJ45 connector and the commercial sold it as a 100Mbit device. But they never had a circuit board capable of moving data at that speed. Which was never a problem because back then not many had fiber with 100/100 speed at home so they never experienced the real roof of the router anyway.
To be honest, if you really want to move 100Mbit data bidirectional continuously and lag free with all the overhead functions that everyone wants in OpenWRT. Then you need the 1:10 ratio and that means Gbit equipment.
Hard to pin down for a whole network, but for an individual link this depends basically on the duty-cycle, so how much of the time a link is busy... Ethernet for example ist either 100% busy (when sending a packet) or 0% when there is nothing to send, by averaging over a few (milli)seconds this on off pattern can be translated into a "load" number, or the percentage of the time span during which the link was busy transferring data. Saturation is clearly defined as 100%, but the question is over what time interval
What he said But just to run a few numbers for relative normal situations:
(fast ethernet, aka 100Mbps gross rate, ethernet encapsulation, IPv4, TCP)
large packets (MTU 1500, the largets ethernet payload size that realistically can be passed over the internet):
100 * ((1500 - 20 - 20)/(1500 + 38)) = 94.93 Mbps
small packets (46 is the smalles payload size for ethernet):
100 * ((46 - 20 - 20)/(46 + 38)) = 7.14 Mbps
So at saturation you will see a goodput (application level data throughput, e.g. in a speedtest or in iperf[2|3]) anywhere between ~94 and 7 Mbps, depending on your packet size mix... Typically applications that want to go fast, use large packets, but some applications, like VoIP use smaller packets, and once you have enough VoIP flows they can also saturate your link, so 95Mbps is a decent first estimate, but keep your application behaviour in mind...
Unless the combined traffic trying to push through the switch exhausts/saturates the switches back-plane capacity... (decent switches have sufficient backplane capacity to make that very unlikely or even impossible)
Yup. And to your point, I don't think I've seen any modern switches that have backplane limitations of note.
Which devices are we talking about? My FastEthernet devices all were capable of reaching the practical speed of ~94Mbps, unless it was some lousy implementation without adequate cpu to handle it.
Fast Ethernet was introduced in 1995 as the IEEE 802.3u standard and remained the fastest version of Ethernet for three years before the introduction of Gigabit Ethernet. The acronym GE/FE is sometimes used for devices supporting both standards.
Just about everything with Linksys on the plastics.