What's new

Wireless router with Jumbo Frame support

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

doubleroll

Regular Contributor
Other than Asus any other recent wireless router have jumbo frame support. I have already run tests on my AC66R and get better throughput with it turned on with both a synology NAS and netgear NAS overLAN of course.


Sent from my iPad using Tapatalk H
 
To check a specific router for that feature is just a google search away.

To know if they will perform similarly or better than the RT-AC66R in your network would require you testing them all individually.
 
Thanks Guys. It seems some or most Manufacturers, Netgear, etc, do not support jumbo frames on their consumer level routers. Guess I can just stay with Asus when I upgrade next...
 
Oh and the strange thing is most manufacturers of these commercial router do not even mention jumbo frames which makes it difficult to google search.

I wonder if say the Netgear R7000 with ddwrt or some other would enable jumbo frames...

Sent from my iPad using Tapatalk HD
 
While some level of jumbo frame support is implicit in some products, I would base my decisions on a firm statement from the manufacturer or third party firmware creator first. Secondly, if you can find evidence on the net that jumbo frames are causing (or not causing) issues with people with similar uses and devices to yours with a specific model, then use that.

But, sticking with what works is what keeps things working. :)
 
How would a consumer router cope with jumbo frames while bridging to DOCSIS or ADSL and onward to the WWW where jumbo-not exists?
 
Most Broadcom SoC's implicitly support Jumbo Frames at the switch on chip fabric level.. so LAN wise, should be fine with most BRCM based router/AP's - other OEM's, I would suspect to be the same - not explicit support, but it should be there.

Just keep in mind that this would be wired only on the LAN ports - so Jumbo's going out to the WLAN or WAN, they're going to get fragmented...
 
Just keep in mind that this would be wired only on the LAN ports - so Jumbo's going out to the WLAN or WAN, they're going to get fragmented...
Will consumer routers actually fragment an incoming LAN-side jumbo packet implicitly addressed to the gateway? Do some just reject it?
 
they will fragment unless the do not fragment is set in which the packet will be dropped. For jumbo frames to work on a networking gear both the hardware and firmware must support it. Some websites provide low level details on what chips a router has and you can see what it supports. The other question than is what the firmware itself supports.

For networks with different frame sizes the CPU that performs the switching/routing/bridging must be fast enough to fragment the packets. When a packet from a smaller MTU goes into a larger MTU it will just be the size of the smaller MTU unless there is some conversion that goes on (such as jumbo frames on layer 2, layer 3 routing over layer 2 using 65KB packets, layer 3 wont fragment but layer 2 will fragment/repack the packet (usually switch level) between hosts). Just because you use jumbo frames doesnt give a performance boost always, all it is doing is increasing the maximum size a packet can be so more data can be packed into a single packet. Think of it as header vs data size efficiency, less header per data, more efficient.

The internet uses 1500bytes (or slightly less) over layer 3 so router support isnt necessary. the cheapest and surest way to add jumbo frames to your network is with a semi managed switch that supports it and having computers with NICs that support it too. The benefit comes from using it over LAN to transfer large data such as files, rendering, huge arrays and so on which reduces CPU usage and latency on said transfers on a CTF based switching.
 
No. No. No.

TCP/IP uses MTU path discovery. All that will happen is that any devices will set their MTU size to the limit that the router will handle. At most you have a few millisecond latency imposed on new TCP connections if you have jumbo frames on. MTU path discovery is extremely fast, so likely the latency on a 1Gbps connection to the router is going to be measured in hundreds of microseconds.

There is NO packet fragmentation going on. Wireshark it, you can see it restrict the packet size after the initial exchange.
 
  • Like
Reactions: tim
No. No. No.

TCP/IP uses MTU path discovery. All that will happen is that any devices will set their MTU size to the limit that the router will handle. At most you have a few millisecond latency imposed on new TCP connections if you have jumbo frames on. MTU path discovery is extremely fast, so likely the latency on a 1Gbps connection to the router is going to be measured in hundreds of microseconds.

There is NO packet fragmentation going on. Wireshark it, you can see it restrict the packet size after the initial exchange.

Beat me to it! Was just about to write about PMTU.

https://en.wikipedia.org/wiki/Path_MTU_Discovery
 
Just keep in mind that this would be wired only on the LAN ports - so Jumbo's going out to the WLAN or WAN, they're going to get fragmented...

Ignore the comment above in most cases... I was speaking more towards edge cases with path discovery returns incorrect results, which does happen..

In general though - the TCP connection negotiates the MTU as part of the handshake... so LAN-LAN might be jumbo, and the same application from LAN to WLAN or LAN to WAN will use an appropriate frame size - so one should be ok - but I've seen brain-dead apps, nics, and switches that have problems..
 
I forgot to add.

The ASUS RT-AC68U original F/W has an option to enable Jumbo Frames. The Netgear R7000 original F/W does not have this option* even thought the hardware is almost identical.


*this was the case a few months back when I did some research into a new AC wireless router.
 
No. No. No.

TCP/IP uses MTU path discovery. All that will happen is that any devices will set their MTU size to the limit that the router will handle. At most you have a few millisecond latency imposed on new TCP connections if you have jumbo frames on. MTU path discovery is extremely fast, so likely the latency on a 1Gbps connection to the router is going to be measured in hundreds of microseconds.

There is NO packet fragmentation going on. Wireshark it, you can see it restrict the packet size after the initial exchange.

For TCP but not every protocol has that. Not sure if you were reading but i mentioned that using jumbo frames does lower latency when CTF is used which is unrelated to routing. Packet fragmentation happens in microseconds in the processor not related to the interface speed. The point is you can create a tunnel between 2 devices that has a higher MTU than layer 2 and it would still work and probably fragment on the lower layers but the end point would just get all the packets and rebuild it in the switch CPU before passing it to the main CPU. Some tunnels are UDP based and some use their own protocol. The network stack basically lets you have a much higher MTU on a higher layer than on the lower layer and vice versa. The only reason to adjust the MTUs to be the same on different layers is for performance reasons since in the past you could still get switches that werent capable of wirespeed.
 
In a lot of cases a lot of adapters only use jumbo frames for TCP, they don't use it for UDP, ICMP, etc. Yes, it is possible to have packet fragmentation and for it to work. The point though is, on a TCP connection, it generally does not happen unless you've explicitly turned MTU path discovery off. It is possible with UDP, but many adapters don't do jumbo with UDP. Or with other protocols.
 

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top