Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: nsp: juniper

Best practice MTU?

 

 

nsp juniper RSS feed   Index | Next | Previous | View Threaded


ObrienH at missouri

Apr 26, 2012, 2:32 PM

Post #1 of 10 (3924 views)
Permalink
Best practice MTU?

We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.

Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.

Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)


Thanks!

Will
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


juniperdude at gmail

Apr 26, 2012, 4:33 PM

Post #2 of 10 (3877 views)
Permalink
Re: Best practice MTU? [In reply to]

I usually set the interface physical MTU as high as it goes (per device), but manually set protocol inet to MTU 1500 (for things like OSPF to work). This allows for as-large-as-MTU-as-MPLS-can-do. Other address families aren't that picky about MTU matching.


ge-1/0/5 {
description "LINK to another IP/OSPF/MPLS device - May or May not support MTU 9192 on the physical.... but inet4/OSPF is 1500 so it works";
mtu 9192;
unit 0 {
family inet {
mtu 1500;
address 10.102.10.1/24;
}
family mpls;
}
}

- CK.



On 2012-04-27, at 7:32 AM, OBrien, Will wrote:

> We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.
>
> Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.
>
> Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)
>
>
> Thanks!
>
> Will
> _______________________________________________
> juniper-nsp mailing list juniper-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/juniper-nsp


_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


jbates at brightok

Apr 26, 2012, 6:42 PM

Post #3 of 10 (3875 views)
Permalink
Re: Best practice MTU? [In reply to]

I agree, as long as the transport between devices supports the MTU. This
is especially important with device interoperability. Cisco, for
example, apparently pads out ISO hello packets to MTU (Juniper limits it
to maximum ISO packet size). If the packet is discarded by transport
medium, the ISIS session will not come up. Found this one out increasing
MTU by 4 to support a single MPLS tag. lol

Jack

On 4/26/2012 6:33 PM, Chris Kawchuk wrote:
> I usually set the interface physical MTU as high as it goes (per device), but manually set protocol inet to MTU 1500 (for things like OSPF to work). This allows for as-large-as-MTU-as-MPLS-can-do. Other address families aren't that picky about MTU matching.
>
>
> ge-1/0/5 {
> description "LINK to another IP/OSPF/MPLS device - May or May not support MTU 9192 on the physical.... but inet4/OSPF is 1500 so it works";
> mtu 9192;
> unit 0 {
> family inet {
> mtu 1500;
> address 10.102.10.1/24;
> }
> family mpls;
> }
> }
>
> - CK.
>
>
>
> On 2012-04-27, at 7:32 AM, OBrien, Will wrote:
>
>> We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.
>>
>> Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.
>>
>> Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)
>>
>>
>> Thanks!
>>
>> Will
>> _______________________________________________
>> juniper-nsp mailing list juniper-nsp [at] puck
>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/juniper-nsp

_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


devinkennedy415 at hotmail

Apr 26, 2012, 7:54 PM

Post #4 of 10 (3892 views)
Permalink
Re: Best practice MTU? [In reply to]

Sent from my Samsung Galaxy Note™, an AT&T LTE smartphone

-------- Original message --------
Subject: Re: [j-nsp] Best practice MTU?
From: Jack Bates <jbates [at] brightok>
To: Chris Kawchuk <juniperdude [at] gmail>
CC: Re: [j-nsp] Best practice MTU?

I agree, as long as the transport between devices supports the MTU. This
is especially important with device interoperability. Cisco, for
example, apparently pads out ISO hello packets to MTU (Juniper limits it
to maximum ISO packet size). If the packet is discarded by transport
medium, the ISIS session will not come up. Found this one out increasing
MTU by 4 to support a single MPLS tag. lol

Jack

On 4/26/2012 6:33 PM, Chris Kawchuk wrote:
> I usually set the interface physical MTU as high as it goes (per device), but manually set protocol inet to MTU 1500 (for things like OSPF to work). This allows for as-large-as-MTU-as-MPLS-can-do. Other address families aren't that picky about MTU matching.
>
>
> ge-1/0/5 {
>      description "LINK to another IP/OSPF/MPLS device - May or May not support MTU 9192 on the physical.... but inet4/OSPF is 1500 so it works";
>      mtu 9192;
>      unit 0 {
>          family inet {
>              mtu 1500;
>              address 10.102.10.1/24;
>          }
>          family mpls;
>      }
> }
>
> - CK.
>
>
>
> On 2012-04-27, at 7:32 AM, OBrien, Will wrote:
>
>> We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.
>>
>> Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.
>>
>> Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)
>>
>>
>> Thanks!
>>
>> Will
>> _______________________________________________
>> juniper-nsp mailing list juniper-nsp [at] puck
>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/juniper-nsp

_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp

_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


sfouant at shortestpathfirst

Apr 26, 2012, 7:59 PM

Post #5 of 10 (3881 views)
Permalink
Re: Best practice MTU? [In reply to]

On 4/26/2012 5:32 PM, OBrien, Will wrote:
>
> Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)

You know before you go about changing MTUs to something over 9000, you
might want to take a look at this youtube video. These two guys talk
about the pros and cons of setting it to over 9000 and after watching it
I kinda get the idea that it might not be such a good thing:

http://www.youtube.com/watch?v=SiMHTK15Pik

--
Stefan Fouant
JNCIE-SEC, JNCIE-SP, JNCIE-ENT, JNCI
Technical Trainer, Juniper Networks

Follow us on Twitter @JuniperEducate
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


saku at ytti

Apr 26, 2012, 10:52 PM

Post #6 of 10 (3865 views)
Permalink
Re: Best practice MTU? [In reply to]

On (2012-04-27 09:33 +1000), Chris Kawchuk wrote:

> I usually set the interface physical MTU as high as it goes (per device), but manually set protocol inet to MTU 1500 (for things like OSPF to work). This allows for as-large-as-MTU-as-MPLS-can-do. Other address families aren't that picky about MTU matching.

+1.

Short of platform specific side-effects it might cause like static
buffering per MTU ranges causing available buffers to be significantly
reduced when enabling large MTU. It is always good idea to set L2 and L3
link MTU to as large as they go, regardless if they match on ends or not,
and then for L3 protocols which rely link MTU being same, set it in
protocol level.
--
++ytti
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


p.mayers at imperial

Apr 27, 2012, 1:56 AM

Post #7 of 10 (3891 views)
Permalink
Re: Best practice MTU? [In reply to]

On 04/27/2012 12:33 AM, Chris Kawchuk wrote:
> I usually set the interface physical MTU as high as it goes (per
> device), but manually set protocol inet to MTU 1500 (for things like

We do almost this (physical -> max), but set IP MTU to 9100 rather than
default 1500. The latter is helpful if you ever want to do tunnels e.g.
mirror traffic to GRE that stay purely within your network.
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


helmwork at ruraltel

Apr 27, 2012, 5:40 AM

Post #8 of 10 (3865 views)
Permalink
Re: Best practice MTU? [In reply to]

On 4/26/2012 4:32 PM, OBrien, Will wrote:
> We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.
>
> Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.
>
> Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)
>
>

I just ran across this the other day. Some good reading and looks like a
push to standardize 9000 byte MTU across the board.

https://ripe64.ripe.net/presentations/139-Hurricane_Electric_-_Martin_Levy_-_IX_Jumbo_Frames_-_RIPE64_EIX-WG_-_Slovenia.pdf

http://www.ietf.org/id/draft-mlevy-ixp-jumboframes-00.txt

/Eric
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


colin at netech

Apr 27, 2012, 7:56 AM

Post #9 of 10 (3864 views)
Permalink
Re: Best practice MTU? [In reply to]

On Fri, Apr 27, 2012 at 07:40:43AM -0500, Eric Helm wrote:
>
>
> On 4/26/2012 4:32 PM, OBrien, Will wrote:
> > We've been pushing out jumbo frames across our new core lately. Right now I've got multiple boxes from multiple vendors that all support different maximum MTUs.
> >
> > Example: Juniper MX960/480, Nexus 7009, Nexus 5k/2k, Catalyst 4900, Nortel/Avaya 8600.... All different maximums.
> >
> > Anyone have suggestions for a best practice MTU? (That is.... over 9000?!)
>
> I just ran across this the other day. Some good reading and looks like a
> push to standardize 9000 byte MTU across the board.
>
> https://ripe64.ripe.net/presentations/139-Hurricane_Electric_-_Martin_Levy_-_IX_Jumbo_Frames_-_RIPE64_EIX-WG_-_Slovenia.pdf
> http://www.ietf.org/id/draft-mlevy-ixp-jumboframes-00.txt

9000 for IP mtu provided to end users / customers is a nice round
number.

I have started using 9100 as the internal mtu as it leaves 100
bytes for any encap overhead you might want from mpls/gre/etc and is
easy to remember.

Colin

--
Colin Whittaker +353 (0)86 8211 965
http://colin.netech.ie colin [at] netech
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp


mark.tinka at seacom

May 20, 2012, 1:30 PM

Post #10 of 10 (3793 views)
Permalink
Re: Best practice MTU? [In reply to]

On Friday, April 27, 2012 04:56:02 PM Colin Whittaker wrote:

> 9000 for IP mtu provided to end users / customers is a
> nice round number.
>
> I have started using 9100 as the internal mtu as it
> leaves 100 bytes for any encap overhead you might want
> from mpls/gre/etc and is easy to remember.

I've always limited vendors to just two, Cisco and Juniper.

As all the Juniper platforms we've worked with top out at
9,192 bytes, that's what we've set both the Juniper and
Cisco kit to (despite the fact that newer Cisco hardware can
go up to 9,216 bytes).

Junos and IOS XR are in sync. re: how MTU is handled.
Junos/IOS XR and IOS/IOS XE are out of sync., so this needs
to be corrected when deploying.

The biggest issue is when you have earlier, low-to-mid range
Cisco switches that can go no higher than 9,000 bytes if
you're looking to have the same MTU value supported edge-to-
edge. Newer mid-range Cisco switches (like the 3560-X/3750-
X) now support 9,216 bytes as well.

Also be mindful of older Cisco 100Mbps Fast-E ports that can
only support 1,500 - 1,988 bytes. Juniper's Fast-E ports
don't have this issue, however, nor do the Fast-E ports that
you can install in Cisco's XR 12000/CRS routers (the SPA's).

Mark.
_______________________________________________
juniper-nsp mailing list juniper-nsp [at] puck
https://puck.nether.net/mailman/listinfo/juniper-nsp

nsp juniper RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.