1

I often do a lot of work from the UK to destination networks in Australia be it voice calls or something more benign like gaming. Both of which work best with the lowest RTT possible.

On average the RTT is around 280ms from my UK network to Australia. The theoretical best RTT, as the crow flies, is far lower. Of course, the internet doesn't work like that -> the pipes are nowhere near straight, packets don't actually go the speed of light, router processing time, etc etc.

This has led me to a curiosity - I've no doubt that by default my route to Australia is not taking the most optimal given that ISPs will BGP peer and route through whatever is cheapest. However I wonder if I can tunnel/VPN through something more direct, IE like an Amazon EC2 server backbone.

Example trace route to www.telstra.com.au:

Tracing route to www.telstra.com.au [203.36.190.11]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  BTHUB5 [192.168.1.254]
  2     *        *        *     Request timed out.
  3     *        *        *     Request timed out.
  4     7 ms     7 ms     7 ms  31.55.186.180
  5     7 ms     7 ms     7 ms  core4-hu0-11-0-3.faraday.ukcore.bt.net [213.121.192.70]
  6     7 ms    11 ms     8 ms  213.121.193.199
  7     7 ms     7 ms     7 ms  t2c3-et-3-3-0-0.uk-lon1.eu.bt.net [166.49.211.238]
  8    10 ms    11 ms    11 ms  i-0-4-0-7-peer.ulco-core02.pr.telstraglobal.net [134.159.95.25]
  9    79 ms    85 ms    76 ms  i-10104.unse-core01.telstraglobal.net [202.84.141.145]
 10   139 ms   138 ms   138 ms  i-1.tlot-core02.telstraglobal.net [202.84.252.85]
 11   138 ms   138 ms   139 ms  i-0-3-0-2.1wlt-core02.telstraglobal.net [202.84.143.202]
 12   285 ms   287 ms   289 ms  i-16.sydp-core03.telstraglobal.net [202.84.136.158]
 13   293 ms   286 ms   288 ms  bundle-ether3.pad-gw10.sydney.telstra.net [203.50.13.85]
 14   285 ms   284 ms   284 ms  bundle-ether2.pad-gw11.sydney.telstra.net [203.50.6.59]
 15   286 ms   291 ms   286 ms  bundle-ether3.ken-core10.sydney.telstra.net [203.50.6.60]
 16   284 ms   284 ms   284 ms  tengigabitethernet7-1.stl1.sydney.telstra.net [203.50.20.50]
 17   283 ms   283 ms   283 ms  telstr1429.lnk.telstra.net [165.228.136.114]
 18     *        *        *     Request timed out.
 19   290 ms   285 ms   284 ms  203.36.190.11

I suppose the question here is: How can I investigate and determine whether I am taking the shortest available (Commercial VPN/tunnel or otherwise) path from the UK to Australia?


Using the live chart over at https://www.cloudping.co/ eu-west-2 (London) <--> ap-southeast-2 (Sydney) has an average RTT of 280 as well. So it appears Amazon's backbone isn't any better than public transit.


Found another test(http://gcp.latenci.es/) for the Google Cloud network and it shows similar results as Amazon - 275 RTT.


Okay. So this is the sub cable map around Australia: https://i.imgur.com/MiBn28m.png

Using an ISP looking glass (https://www.us.ntt.net/support/looking-glass/) I ran some tests from various routers. No matter what, all traffic is routed via Singapore, and from Singapore to Australia which has a very abnormally high RTT. Even running a route test from a West USA (LA) router to Sydney routes it via Singapore!

It does not appear that any publically available routes are going via the US West <--> Sydney sub cables. And it certainly looks like the Singapore <--> Australia route is being smashed.

closed as off-topic by Ron Maupin Sep 8 '18 at 15:20

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "NE is a site for to ask and provide answers about professionally managed networks in a business environment. Your question falls outside the areas our community decided are on topic. Please visit the help center for more details. If you disagree with this closure, please ask on Network Engineering Meta." – Ron Maupin
If this question can be reworded to fit the rules in the help center, please edit the question.

  • You must also consider congestion at each point along the route. This map is interesting: submarinecablemap.com – Ron Maupin Sep 8 '18 at 14:43
  • Indeed. I'm starting to think almost everyone comes in via Singapore, and the SP -> Sydney hop/s add a huge amount of latency for the comparably small distance over the greater path. – S.Richmond Sep 8 '18 at 14:46
  • You have changed the question to, "Why is access to Australia so high latency?" That question is off-topic here as a question about a network you do not directly control. It is not a question we can answer. Your original question was also basically off-topic. I really do not understand how we can answer your question or help you. – Ron Maupin Sep 8 '18 at 15:17
  • Reworded the question - Surely someone can assist with figuring out whether there is an optimal route to take that is within the control of a potentially commercial solution? – S.Richmond Sep 8 '18 at 15:23
  • Now, you are asking, "How can I reduce latency to Australia from another country?" Unless you can give us a good description of the network, the network device models, and the network device configurations for the path, we cannot help. You do not directly control the networks in the path, which is a requirement to be on-topic here. If you are looking for a recommendation, the that is off-topic for SE sites, except Software Recommendations and Hardware Recommendations. Your question could make an interesting discussion on Network Engineering Chat, and you have enough reputation to participate there. – Ron Maupin Sep 8 '18 at 15:26
1

Calculate the theoretical lower bound by great circle distance.

  • London to Sydney is 17,000 km (gcmap)
  • Speed of light is 300,000 km/sec
  • Velocity factor of fibre 0.7
  • Minimum RTT is 2 * (17/300) / 0.7 = 162 ms

The difference between a great circle distance and the real fibre routes is not usually terribly important in comparison with number of routers so on.

For typical best engineering practice, use ping times from one AWS datacentre to another in the locations you care about.

For a recent live problem for one of my clients, they wanted to know if the ISP was giving them good lag between London and Singapore.

  • Lower bound 11,000 km = 105 ms
  • AWS London to Singapore 165 ms
  • Client's ISP average 342 ms
  • Other client's ISP 192 ms

We are still arguing with this client's ISP about whether its service is reasonable. And for temporary improvement, this client is doing exactly as you suggest and routing a VPN through AWS networking.

  • Cheers. I just found a live chart (cloudping.co) of the Amazon backbone showing RTTs. Looks like London <--> Sydney is actually no better than public transit, weirdly. – S.Richmond Sep 8 '18 at 14:26
  • I find that surprising too. My numbers came from servers I launched in London and Singapore so I could do lots of tests as well as simply ping; sadly cloudping doesn't have numbers for Singapore. – jonathanjo Sep 8 '18 at 14:30
  • Found a test for Google's cloud platform too (Edited in OP) - Same results! :/ – S.Richmond Sep 8 '18 at 14:38

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