IPv4 are 32 bit addresses, while IPv6 are 128 bit. It's 4 times larger, sure on a small scale maybe not a big difference, but on a world wide scale will it make a difference?

Are their any messures in place to reduce the load? Caching, or algorithms that make lookups faster, or so your transfer less with each packet?

  • 1
    Oddly enough, we tried to debate that having too much memory would slow your computer down by increasing the size of the lookup table. That was good for laughs. – surfasb Mar 24 '12 at 1:40
  • 2
    The IPv6 address space is not 4 times larger, it's 2^96 times larger. – Michael Hampton Sep 13 '12 at 11:00

Here is the differences between IPv4 and IPv6 that can affect performance:

  • IPv6 routers do not perform fragmentation - Path MTU discovery is required and the upper layers are expected to not send packets larger than the MTU
  • IPv6 header doesn't use a checksum
  • IPv6 header is simplified - the main part of header is fixed length, with the provision for "extension" headers
  • Supports jumbo payloads - theoretically this can increase performance by decreasing the amount of transmission dedicated to sending headers

Of course router, data link, and physical layer implementation all play a role too.


There might be a difference once you push your equipment to its limits. Then the very little extra overhead that IPv6 has might be measurable. But most equipment will not be used constantly at its limits, certainly not on global infrastructure. So for most practical purposes IPv6 will be equally fast as IPv4.

In the (near) future when regions run out of IPv4 addresses this will change. Then extra NAT will be needed even at an ISP level (either directly as Carrier-Grade-NAT/NAT444 or combined with IPv4-over-IPv6 tunneling in DS-Lite), and that will cause overhead for handling IPv4. Then IPv6 might become noticeably faster than IPv4 because packets can travel over the network without needing processing/changing in central NAT routers.


This IPv6 Land Speed Record documentation has some nice numbers:

In the same period and with almost the same setup (the distance was shorter, but the endstations were the same) also the TCP/IPv4 LSR was set. But IPv4 was better:


Distance: 10949 KM
Data transferred: 2.3 Terabytes
Average speed (over 60 minutes): 5.64 Gbps
Record submitted: 61,752,360,000,000,000 meters-bits/sec


Distance: 11539 km
Data transferred: 560 Gbytes
Average speed (over 20 minutes): 4.00 Gbps
Record submitted: 46,156,000,000,000,000 meters-bits/sec

But I would assume, globally, the result may be different due to the IPv6 jumbograms:

IPv4 limits packets to 65535 (216−1) octets of payload. An IPv6 node can optionally handle packets over this limit, referred to as jumbograms, which can be as large as 4294967295 (232−1) octets.

So, even though the address is longer, the payload is much bigger.

  • 1
    I don't think land-speed-records are relevant for the question asked because they use very special setups/configurations. – Sander Steffann Mar 22 '12 at 19:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.