Hi!
This is the ever present confusion on MTU.
You (and most of the rest of the world with you) needs to distinguish between L2 and L3 MTU (not saying that you don't, just that you need to). L2 MTU is the Ethernet FRAME size. This can normally be as large as the switch allows. It tells the switch how big Ethernet FRAMES to allow on the receive side of things (and transmit). This is normally referred to as jumbo frames. You can enable jumbo frames on all switches with very little risk of messing something up.
It is the endpoints and/or routers that determine the L3 MTU. This is the IP PACKET size. I emphasize the difference between frame and packet as they are commonly confused for each other. Of course, the L2 MTU can differ if using another medium, like ATM, wireless of different kinds and so on, but I assume Ethernet here.
If the Ethernet frame L2 MTU is set to 1514 bytes, you can fit a 1500 byte IP packet in there if you don't have VLAN tags (1518 if you have one tag).
It is when you start using a higher L3/IP MTU larger than 1500 that things can go wrong. If you want to use larger packet sizes for increased throughput locally and still use 1500 bytes for Internet bound traffic, your (or your ISPs) border router needs to fragment outbound packets for you. That is not an issue for most modern routers, but you might need to tell the router to do so. Packets from the Internet will never have an L3 MTU of more than 1500 so reassembly will almost never occur (well, the national university interconnects may be a different story).
The performance gain may vary from a few percent, which is the overhead in terms of Inter Frame Gap, preable, Ethernet and IP headers and so on being cut, but application wise, I've seen tests with 110 % performance gain, but also 5 % performance loss when using a 9000 byte L3 MTU. Firewalls tend to like big packets to process, so perfformance in an SRX may increase a lot if enabling a higher IP MTU throughout your network.
Hope this helps rather than the opposite :)
Original Message:
Sent: 12-07-2020 08:13
From: Unknown User
Subject: Is it MTU different between use family bridge vs family inet on MX480?
Hi,
I'm already check the "interface extensive" and we can see the MTU error increase. The weird thing the MTU irb interface that involve in MC-LAG (vrrp) the MTU is not follow the physical interface if i increase the value of mtu physical interface. If the the irb not in MC-LAG interface then it will automatically follow the MTU physical interface incase i increase the MTU.
After i increase the irb MTU (vrrp) in MC-LAG then i can see the mtu error stop. For example "ae1" mtu 9192 then irb.x mtu 9178.
https://kb.juniper.net/InfoCenter/index?page=content&id=KB24601&cat=MX_SERIES&actp=LIST
Thanks and appreciate any feedback
Original Message:
Sent: 12-01-2020 06:53
From: Unknown User
Subject: Is it MTU different between use family bridge vs family inet on MX480?
Hi smicker,
I'm using family bridge coz MC-LAG. May i know why i still need to add vlan-tagging? Coz i'm simulate on vmx and it have different result when using family bridge and normal vlan-tagging.
Thanks and appreciate your feedback.
Original Message:
Sent: 12-01-2020 02:25
From: Unknown User
Subject: Is it MTU different between use family bridge vs family inet on MX480?
MTU is not increased automatically on MX/EX92xx L2 trunked interfaces, you will need to add vlan-tagging.
https://kb.juniper.net/InfoCenter/index?page=content&id=KB25421&cat=MX_SERIES&actp=LIST
Original Message:
Sent: 11-29-2020 21:02
From: Unknown User
Subject: Is it MTU different between use family bridge vs family inet on MX480?
Hi all,
Currently my setup as below. Between CE --> MX i'm just use default MTU. When i change setup using family bridge on MX480 (irb.70) then i cannot ping when do command "ping CE-IP-ADRESS size 1500". But when i change on MX480 using interface ge-0/0/0 unit 70 then i can ping with size 1500. Is it when do family bridge then need to increase interface on MX480 to 1518 instead by default just 1514?
CE ---> EX4300 ----> MX480
Thanks and appreciate any feedback