Breaking 100K Entries is the Global IPv6 BGP Table

This year, 2020, around November I started to see the global IPv6 BGP Table is getting more than 100K entries. Although the number is going above and under 100K from time to time, starting from the end of November I can safely say it is breaking 100K entries right now.

This is an interesting milestone for IPv6. That means a massive majority of people are using IPv6 today. I want to note down this moment. And I want to share 3 of my own observations about the IPv6 BGP table.

Source: https://twitter.com/bgp6_table/status/1330964625127583744/photo/1

Number of IPv6 BGP entries is going up and down

Internet is a collection of distributed, self-managed networks. No single authority can dictate how the BGP configuration should be done on all different networks. Each of the network administrators could choose when and how to add or remove BGP entries in different occasions.

It is natural for BGP table to grow and shrink from time to time.

For example, network administrators might decide to remove assigned but not used yet networks from BGP configurations. By this action, the number of BGP entries could go down.

Another example, to achieve load spreading, administrators might break their own IPv6 prefixes into smaller ones and advertise them to different BGP neighbors. By this action, the number of BGP entries could go up.

And of course, upon expanding networks with more prefixes, or dying companies returning prefixes to Internet Regional Registries, the BGP table would grow or shrink accordingly.

Configuration errors could also result in fluctuations of the total number of BGP entries.

Source: IPv6 CIDR Report for 30 Nov 20

Projection of 100K time is pretty accurate

I have read web pages by APNIC and RIPE projecting the total number of IPv6 BGP entries. They all projected the time of 100K is around the second half of year 2020. 

They are pretty accurate in my opinion.

Source: APNIC, "BGP in 2019 - The BGP Table"

Source: RIPE, "BGP in 2016"

IPv6 BGP table is growing much faster than IPv4

For IPv6 BGP: Last year, 2019, around October, I observed IPv6 BGP table is around 80K in size. After around 1 year now, it is over 100K. That is, the growth rate in this interval is 25% (=20/80).

For IPv4 BGP: same interval as above, October 2019, I observed IPv4 BGP table is around 800K in size. After the same 1 year up to now, it is over 850K in size. That is, the growth rate in this interval is 6.25% (=50/800)

My conclusion is: the growth rate of IPv6 is much higher than that of IPv4, in this interval.

One more thing…

Many people are also interested in estimates of router memory consumption to hold the whole global BGP table. For IPv6, unfortunately, I cannot find good firsthand samples about the number of entries versus the memory size consumed. I now try to estimate the memory consumption by samples of IPv4.

For single IP address, IPv4 is 32 bits, and IPv6 is 128 bits. One IPv6 address is 4 times the size of one IPv4 address. Because essentially BGP entry fields are IP addresses, here I roughly assume IPv6 BGP table should not take more than 4 times the memory consumption of IPv4 BGP table of the same number of entries.

I already wrote about this before: every 100K IPv4 BGP entries could take no more than 80 Megabytes of memory. Therefore, my estimate for the same 100K entries of IPv6 BGP table, should not take more than 320 Megabytes of memory.

Do you have firsthand numbers of IPv6 BGP memory consumption? How wrong is my estimate ? I would like to hear from you in the comment section below.

Overlooking from the top floor of Dragon and Tiger Pagodas (龍虎塔)
Zuoying District, Kaohsiung City, Taiwan

I am Li-Ji Hong. This is my blog “Show IP Protocols”. See you next time!