Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Pages: 1 [2]

Author Topic: Tweaking speed out of my ECI cabinet  (Read 3084 times)

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: Tweaking speed out of my ECI cabinet
« Reply #15 on: September 29, 2020, 09:40:17 AM »

Even though I have an IPv4 MTU of 1500 bytes achieved through PPPoE 1508 byte baby jumbo frames, we have to remember that as far as IP is concerned these numbers are just arbitrary. Some IPv6 applications, iirc, just stick to 1280 bytes because that is guaranteed to work in the IPv6 standard, and don’t bother with trying to get up to the path MTU. The number 1500 only comes from ethernet, and has nothing to do with IPv4 or IPv6. Having said all that, I wanted a 1500 byte IPv4 MTU for largely illogical reasons, wanted best byte overhead efficiency, and best compatibility with IPv4 applications. For IPv6, for reasons that have been discussed in earlier threads, I use an MTU of 1408 though.
Logged

Robbie

  • Member
  • **
  • Posts: 75
Re: Tweaking speed out of my ECI cabinet
« Reply #16 on: September 29, 2020, 01:41:02 PM »

Things are getting better with ipv6 in recent times and I don't constrain it now.  I also run with scissors.

Quote
Frame 98: 1514 bytes on wire (12112 bits), 1514 bytes captured (12112 bits) on interface en0, id 0
Internet Protocol Version 6, Src: ######.net (2a04:::###), Dst: 2a02:390:8641:3:cd5a:::#### (2a02:::####)
Payload Length:  1460
Transmission Control Protocol, Src Port: 80, Dst Port: 49285

I didn't read how you got to an MTU of 1408 but it looks like a pretty good number to balance between performance vs avoiding pmtud / icmp pitfalls, so I totally get it.

Cloudflare did a blog on the subject a couple of years ago - for those that may have missed it:

https://blog.cloudflare.com/increasing-ipv6-mtu/

I should probably add that my MTU interaction with QoS is probably down to me using HTB+FQ_Codel.  I understand that the later generation (eg CAKE) does not have this hesitation, but I have yet to try.

👍
Logged

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Re: Tweaking speed out of my ECI cabinet
« Reply #17 on: September 29, 2020, 02:18:44 PM »

See :
    https://forum.kitz.co.uk/index.php/topic,22812.msg387967.html#msg387967
    https://forum.kitz.co.uk/index.php/topic,22630.msg385743.html#msg385743
for a very boring story.

MTU 1408 + overhead is a multiple of 48 bytes, so it is an efficient choice, but it’s not my choice because of its efficiency- it’s only chosen because I have to live with reduced MTU.
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 7382
  • VM Gig1 - AAISP L2TP
Re: Tweaking speed out of my ECI cabinet
« Reply #18 on: September 30, 2020, 11:10:16 PM »

Things are getting better with ipv6 in recent times and I don't constrain it now.  I also run with scissors.

I didn't read how you got to an MTU of 1408 but it looks like a pretty good number to balance between performance vs avoiding pmtud / icmp pitfalls, so I totally get it.

Cloudflare did a blog on the subject a couple of years ago - for those that may have missed it:

https://blog.cloudflare.com/increasing-ipv6-mtu/

I should probably add that my MTU interaction with QoS is probably down to me using HTB+FQ_Codel.  I understand that the later generation (eg CAKE) does not have this hesitation, but I have yet to try.

👍
they back on 1280 now, there is a more recent blog entry where they discovered 1500 was breaking things, my suggestion for ipv6 is 1280.
Logged
Pages: 1 [2]