Poor route to Singapore server using Warp, and increased latency.
What steps have you taken to resolve the issue?
I live in Australia and use the Warp and when I try to connect to a server in Singapore, I see high latency.
Here are the tracert details:
Tracing route to 164.52.28.62 over a maximum of 30 hops
1 7 ms 7 ms 11 ms 172.68.209.86
2 5 ms 4 ms 4 ms 104.28.0.0
3 * * * Request timed out.
4 7 ms 5 ms 5 ms 108.162.247.82
5 101 ms 104 ms 105 ms 198.41.140.222
6 146 ms 131 ms 134 ms 198.41.160.97
7 147 ms 144 ms 145 ms 198.41.140.52
8 174 ms 161 ms 160 ms 121-222-163-203-static.tpix.net.tw [203.163.222.121]
9 162 ms 160 ms 160 ms 10.55.208.133
10 158 ms 164 ms 184 ms 1.9.241.78 (AS4788)
11 * * * Request timed out.
12 161 ms 157 ms 159 ms 164.52.72.9
13 189 ms 159 ms 158 ms 164.52.28.62
Trace complete.
According to the information I found on Peeringdb, the IP(AS63199) I want to connect to has a peer in Singapore, and Cloudflare is also peering here. But they still routing to Taiwan and peering with TPIX then transit to a regional ISP(AS4788) to the destination.
Is there a way to get them to use a more direct route?
The surprising thing is that the traffic routes via Singapore and Hong Kong to reach Taiwan on Cloudflare’s backbone.
Tracing route to 164.52.28.62 over a maximum of 30 hops
1 7 ms 7 ms 11 ms 172.68.209.86
2 5 ms 4 ms 4 ms 104.28.0.0
3 * * * Request timed out.
4 7 ms 5 ms 5 ms 108.162.247.82
5 101 ms 104 ms 105 ms 198.41.140.222(Singapore)
6 146 ms 131 ms 134 ms 198.41.160.97(Hong Kong)
7 147 ms 144 ms 145 ms 198.41.140.52(Taiwan)
8 174 ms 161 ms 160 ms 121-222-163-203-static.tpix.net.tw [203.163.222.121]
9 162 ms 160 ms 160 ms 10.55.208.133
10 158 ms 164 ms 184 ms 1.9.241.78 (AS4788)
11 * * * Request timed out.
12 161 ms 157 ms 159 ms 164.52.72.9
13 189 ms 159 ms 158 ms 164.52.28.62
I ran an MTR from Warp (Mumbai server) to the same IP, and it also routes to Taiwan via Singapore.
Additionally, I ran an MTR from Sydney to another IP, 118.193.18.201, belonging to the same AS63199 - CDS Global Cloud Co, located in Singapore. This traffic is routed directly.
I checked these two IPs from AS63199 in HE bgp toolkit, and they have exactly the same propagation graph, so I’m more inclined to think that this is Cloudflare’s issue.
I hope their administrators can see this and solve the problem.
I sent an email to Cloudflare NOC yesterday, and they responded to me quickly and temporarily changed the route to go through StarHub (AS4657) in SG and that is a more direct route.
As far as I know, in addition to its own backbone network, Cloudflare also has several local carriers as its upstream in Australia, including Vocus (AS4826) and Telstra (AS1221).
Telstra’s routing to East Asia has always been very strange, and often requires a long detour through the United States. I have emailed their NOC but have never received a response.
So I still think it’s a Cloudflare issue, because they fixed it before, but now it’s getting worse.
I emailed Cloudflare’s NOC again and they responded almost immediately, thanking them for their hard work!
This is the route now:
Tracing route to 164.52.28.62 over a maximum of 30 hops
1 5 ms 4 ms 4 ms 162.158.168.18
2 4 ms 9 ms 5 ms 104.28.0.0
3 19 ms 4 ms 4 ms 162.158.168.1
4 9 ms 6 ms 19 ms 108.162.247.84
5 * * * Request timed out.
6 99 ms 98 ms 99 ms be2546.ccr31.sin01.atlas.cogentco.com [154.54.1.29]
7 100 ms 99 ms 98 ms 154.18.2.202
8 * * * Request timed out.
9 105 ms 104 ms 99 ms 164.52.72.13
10 100 ms 99 ms 99 ms 164.52.28.62
Trace complete.
What a nice route! It leads directly to Singapore. They seem to have directly modified the routes of all AS63199 Singapore IPs, and now all routes go through AS174.