Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

What is the best Amazon AWS region to host a website for an Australian audience?

Amazon's marketing material seems to suggest that Singapore is the best place, but the ping tests I've performed have been inconclusive. One test came up with ~220ms ping times for California and ~400ms for Singapore, another test returned ~250ms for both sites.

Does it even matter than much?

share|improve this question

4 Answers 4

I would suggest that the answer is now Sydney, Australia!

http://aws.amazon.com/about-aws/whats-new/2012/11/12/announcing-the-aws-asia-pacific-sydney-region/

Ping times from a residential ADSL2 connection in Melbourne:

$ping -c 10 ec2-54-252-25-79.ap-southeast-2.compute.amazonaws.com
PING ec2-54-252-25-79.ap-southeast-2.compute.amazonaws.com (54.252.25.79): 56 data bytes
64 bytes from 54.252.25.79: icmp_seq=0 ttl=49 time=29.044 ms
64 bytes from 54.252.25.79: icmp_seq=1 ttl=49 time=30.983 ms
64 bytes from 54.252.25.79: icmp_seq=2 ttl=49 time=29.424 ms
64 bytes from 54.252.25.79: icmp_seq=3 ttl=49 time=30.563 ms
64 bytes from 54.252.25.79: icmp_seq=4 ttl=49 time=31.842 ms
64 bytes from 54.252.25.79: icmp_seq=5 ttl=49 time=31.564 ms
64 bytes from 54.252.25.79: icmp_seq=6 ttl=49 time=30.730 ms
64 bytes from 54.252.25.79: icmp_seq=7 ttl=49 time=30.797 ms
64 bytes from 54.252.25.79: icmp_seq=8 ttl=49 time=32.737 ms
64 bytes from 54.252.25.79: icmp_seq=9 ttl=49 time=30.984 ms

--- ec2-54-252-25-79.ap-southeast-2.compute.amazonaws.com ping statistics ---
10 packets transmitted, 10 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 29.044/30.867/32.737/1.024 ms

Sweeet!

share|improve this answer
1  
Yeah, I think we have a nice acceptable answer now. Heh. –  ceejayoz Nov 13 '12 at 14:10

NOTE: I don't know why this got downvoted. To me the general question is how to choose a good location that is near (for certain definitions of near) your main audience.

In my opinion the distance that is mainly important is hop-count. If you find a route from your location to your server that has 3 hops vs a location that has 20 hops the choice should be clear.

Of course routing stuff may change and just choosing by using only your provider isn't quite enough. I'd get accounts from the majority of large ISPs and do some tests.

What test? ping times and hop counts. By ping tests I don't mean do a one time test with 4 pings. Let it run a week or so and do the math. (min/max/avg, standard deviation, packet loss etc. -- yes ping will tell you, just be sure to save the data somewhere in case you find a question that ping won't tell you out of the box) That should give you at least a basic idea of how wether you'll have to expect slowness because of the location.

Hop count is the same. Let it run for some time and do the math again.

With that being said: If you don't know wether it'll matter that much you probably shouldn't care about it. If you know that it'll matter how good the link is between your main audience and yourself I have a feeling that EC2 may not be the best choice in the first place. I don't think you'll get any support if you find that the performance is lacking. I'd rather talk to a service provider that is more open to customer requests and that will respond timely for exactly these kinds of problems.

On another note: Do you really know that your audience is in a certain location or is this just an "australian site" because the people who make it are from australia?

share|improve this answer
    
Perhaps you should look up the definition of "disclaimer". –  John Gardeniers Jun 8 '11 at 12:35
    
Right. Not a native speaker and my brain had a totally different meaning stored under disclaimer than what it actually is -- corrected (hopefully) –  Server Horror Jun 8 '11 at 16:56
1  
Reading your answer I would not have guessed English is not your native language. Good job. –  John Gardeniers Jun 8 '11 at 22:19
    
I can't ping AWS machines directly, so testing was performed by configuring identical web servers (MS Server 2008 + IIS 64bit AMI on t1.micro) in each region and hitting them with a website load testing tool (WCAF) for 2 minutes each. The times referenced are not strictly ping times, but are "time to first byte" times. I tried to skip some of the details to get to the gist of it. PS Definitely Australia only; the products / logistics are Aus only. –  Daniel Crowe Jun 9 '11 at 4:03
    
Just out of curiosity: you can't ping them because the Windows Firewall is activated or because Amazon filters ICMP entirely? And check on the path latency - that might get intresting. –  the-wabbit Jun 9 '11 at 8:32

I agree with the gist of this question. I am located in Australia and I was using AWS with a Web Console for a server located on the East Coast region.
As the Console used AJAX extensively, network latency made the console unusable.

share|improve this answer
1  
Good point about AJAX becoming unusable with high latencies. –  mr-euro Apr 10 '12 at 11:05

The geographic location would really not matter that much. Considered the fact that any data would be at most roughly going 20.000 km in one direction and the same length of wire back (quadruple that when using a sattelite link), you would get a maximal "geographical" latency due to the maximum speed of light of around 120 ms if you were accessing servers in Greenland. What's usually more important is the latency of the network equipment and infrastructure which can quickly outgrow the geographical latency.

Your measured 250 ms is a rather high RTT. Given the geographic distance from Canberra to Singapore (roughly 6,000 km) so you should see something that mainly is the latency of the network equipment - with decent links and not too many hops you should be in the range of 50 - 100 ms. Your measured values might be a hint that you are using a connection going over a high-latency link somewhere or your data is simply passing too many hops.

As there are plenty of sea cables from Australia to the Asia-Pacific zone, so there should be no need for that. Check your data path using the traceroute or tracepath utilities to get a more complete picture.

But even latencies in the magnitude 200 ms should not present significant problems for a web site - people are using mobile cell networks for browsing where latencies may vary between 100 and 4000 ms and are happy with it. If you have something that does intensive processing, the processing time quickly might get more of a concern than network latency.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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