Somehow pings to my VPS in SG (Vultr) is significantly worse over ipv6 compared to ipv4. Anyone know whats up with that?
From 13ms to 206ms. I feel like something is off with their routing. Like, is it being routed out to US or something before being bounced back to SG?

IPv6


IPv4


This post has been edited by ruifung: Nov 21 2023, 03:17 PM

Same here, sluggish on IPV6.

https://forum.lowyat.net/index.php?showtopi...ost&p=108583140

How'd you even get speedtest to work on IPv6 specifically?

For now I'm getting around their ipv6 wonkiness by routing my ipv6 traffic out through a VPS.
I got a vultr (SG) vps set up for ipv6 tunneling because the backup 4G stick don't do IPv6...


This post has been edited by ruifung: Nov 22 2023, 05:40 PM

nothing special, just enable IPV6 PPPoE --> toggle wifi off and on --> speedtest --> check if IP is latched on ipv6.

Somehow it's routing though NTT Japan, NTT America, "Tinet GmbH", before going back to vultr in singapore. It's really going on a trip around the world....
Tracing route to 2001:19f0:4400:4001:5400:ff:fe32:b7e5 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms <REDACTED>
2 * * 7 ms 2001:d08:b1:2::2
3 * * * Request timed out.
4 14 ms 14 ms 12 ms 2001:d08:b1:22::6
5 6 ms 7 ms 7 ms 2001:218:e000:5000::3f9
6 27 ms 19 ms 7 ms ae-5.r23.kslrml02.my.bb.gin.ntt.net [2001:218:0:2000::1b2] <------------- NTT Japan
7 * * * Request timed out.
8 90 ms * 93 ms ae-13.r33.tokyjp05.jp.bb.gin.ntt.net [2001:218:0:2000::c6]
9 * 177 ms * ae-4.r25.lsanca07.us.bb.gin.ntt.net [2001:218:0:2000::62]
10 174 ms 186 ms 181 ms 2001:418:0:2::212 <---------------------------------------------------------- NTT America
11 176 ms 178 ms 181 ms 2001:418:0:4000::9e
12 319 ms 320 ms 328 ms 2001:668:0:2:ffff:0:8d88:6e9e <--------------------------------------------- Tinet GmbH
13 320 ms 327 ms 320 ms 2001:668:0:3:ffff:3:0:46
14 218 ms 217 ms 217 ms ethernetet-0-0-35-sr1.sgp1.constant.com [2001:19f0:fc00::a4f:1a1] <--- Vultr
15 320 ms 320 ms 321 ms ethernetswp55-ds1-u2-1155-a.sgp1.constant.com [2001:19f0:fc00::a4f:136]
16 * * * Request timed out.
17 320 ms 319 ms 320 ms 2001:19f0:4400:4001:5400:ff:fe32:b7e5

Trace complete.

Only Maxis can fix this problem. It is because they didn't peer with Equinix SG on IPv6.
Here is how to verify BGP routing...

Refer to Maxis IPv4 peer:
https://bgp.he.net/AS9534#_peers

You can see from your IPv4 traceroute it go from Maxis directly to Equinix SG.

From https://bgp.he.net/AS9534#_peers6 you can see Equinix SG is not peered for IPv6.
Alternatively, you can see the IPv6 don't have "X" under Peer v4.

You then need to check Graph v6: https://bgp.he.net/AS9534#_graph6
Line in BOLD means they are upstream provider.

Maxis has 3 upstream provider: Telecom Italia, NTT, TATA.

So Maxis just dump the packet to one of their upstream, which is NTT.
NTT route the packet to AS3267, which is GTT Communications.
It is not TiNet.

GTT has a public Looking Glass:
https://www.as3257.net/lg/

2001:19f0:fc00::a4f:252 is located in London, so you can choose any LON router, and enter your destination address:
2001:19f0:4400:4001:5400:ff:fe32:b7e5

You can watch your packet traverse their MPLS network.

You can also do the reverse and start from https://bgp.he.net/AS20473#_peers6
You might be wondering if both the source and destination share a common peer, which is NTT, why do the packet need to traverse GTT?

Fortunately, NTT also has a Looking Glass:
https://www.gin.ntt.net/looking-glass-landing/

If you select BGP and see the prefix announcement, the two network didn't actually connect directly

This post has been edited by kwss: Nov 27 2023, 06:08 PM

Based on this Maxis route are better then tmnet?

Looking at this one specific traceroute, TM is better

So every traceroute the results will be different?

There are a lot of quirks in traceroute. Someone wrote about it here:
https://gekk.info/articles/traceroute.htm

Remember to read the SlideShare about it too, linked here:
https://www.slideshare.net/RichardSteenberg...with-traceroute

A general question I see people ask before they sign-up is which ISP has better route?
This question don't have an answer. It is like asking which ISP build better roads. There are a lot of different roads in the Internet. Some are better, some are worse. What matter is the road you use often and the time of day you use them.

Even then, road change often with peering, Policy Based Routing, congestion, which may or may not have anything to do with your ISP at all.

I just give a hypothetical example for this specific case:
If Vultr were to be in SGIX instead of Equinix SG, it might work best on Maxis and worst on TM.

Like I pointed out above, the best thing for consumer is for each ISP to have a public Looking Glass portal. This way we can check out how each ISP route before signing up. Gamer will use the Looking Glass to check for their game servers, during the time when they play game.