Route trace from Beijing Unicom Residential network, looks like packets are routed through japan, not sure if it's just temporary or will be improved later. (Comparing 50+ms from Beijing to other providers in Korea):
3 5 ms 6 ms 3 ms 61.148.160.169
4 8 ms 7 ms 6 ms 124.65.57.117
5 6 ms 7 ms 6 ms 123.126.0.77
6 36 ms 38 ms 39 ms 219.158.7.22
7 109 ms 103 ms 105 ms 219.158.23.10
8 42 ms 39 ms 41 ms 219.158.19.81
9 165 ms 165 ms 163 ms 219.158.33.22
10 89 ms 88 ms 87 ms 106.187.6.173
11 164 ms 165 ms 168 ms 118.155.197.178
12 177 ms 177 ms 177 ms 106.187.28.90
13 94 ms 97 ms 100 ms 27.0.0.228
14 * * * Request timed out.
15 179 ms 182 ms 180 ms 54.239.123.136
16 95 ms 94 ms 93 ms 54.239.123.125
17 98 ms 133 ms 102 ms 54.239.122.226
18 161 ms 152 ms 143 ms 54.239.122.241
19 101 ms 102 ms 99 ms 54.239.122.6
20 * * * Request timed out.
21 * * * Request timed out.
22 * * * Request timed out.
23 100 ms 99 ms 100 ms my.vps.ip.addr
Even more sad, according to our customers and sales leads, is the lack of a Canadian presence. It's getting harder and harder to believe that they have nothing in Canada.
I think we[1] are going to open Montreal this quarter[2]. Not much help if you need EC2 style services, but if you just need to store some data in Canada ... you'll have an option. Drive-up service too.[3]
What is the killer app for a hosting region in Africa? All I have seen is gamers in South Africa complaining about the ping times to servers located in EU making real-time games not worth playing for them. But that's a relatively niche market.
When I was at AWS I visited South Africa for a conference, and paid a visit to the local office in Cape Town. I don't think I can disclose any other detail (number of people, etc), but it is public knowledge that there is a substantial workforce still there, doing development and support.
The problem with an AWS region in Africa is, in my view, that there's not a large market to support the idea. South Korea is for sure more interesting.
Meanwhile, Google Compute Engine still only has four regions and hasn't expanded in over 18 months. (Don't get me started on App Engine, it's not even available for the Asia region!)
Still no presence in the entire southern hemisphere!
AWS have at least 32 DCs - all regions have a minimum of two data centres for proper ha. There are four more scheduled to come online this year, India and UK.
If you think ovh and aws are comparable in anything but the very broadest terms you've not understood what AWS are offering.
OVH currently has 15 DC, of those additional 12, most of them are in new region, Australia, Singapore, Germany, US, and I think Japan. That will make 27 DC.
This isn't say they are comparable by size. Not at all. But you make it sounds like OVH is a small player. Linode, Digital Ocean are medium size. According to Netcraft AWS being first, DO recently became second, and OVH being third. But with the lots of $5 - $20VM DO host, those are VM numbers, compared to the Server Size OVH host.
I would also add that since OVH own and operate a large physical network[0], they can't really be compared to a hosting-only provider like DigitalOcean. This isn't 'like for like', and is only marginally useful for ranking either company.
Also, for those of us looking for server space to serve a primarily Western market, OVH and AWS are most definitely comparable. AWS have a huge and impressive range of features, most of which I simply do not need (though I'm a happy OVH, AWS, and DO customer because I'm not one to put all my eggs in one basket).
Not sure if everyone realizes this, but often (outside the US at least) neither company build or owns data centers. For example, Amazon Sydney is in Equinix SY3[1].
I'm guessing Equinix aren't allowed to say AWS is in their DC.
But their equipment is in there, 100% guaranteed (or at least was in 2012).
Edit: Here's a better reference:
Equinix yesterday launched phase two of its Sydney 3 International Business Exchange datacentre facility (SY3-II) in Alexandria, with Amazon Web Services (AWS) onboard as one of its first customers.... It was only confirmed yesterday — despite it being the worst-kept secret in the industry - that it is using Equinix's SY3-II datacentre.
AWS have two dedicated DCs ('AZs') in Sydney. For services that require quorum (Dynamo, SQS, S3 and others) they also have a private third 'AZ' in SY3-II. When you get an EC2 instance it is always in a dedicated DC, not Equinix.
I don't know about OVH. I know OVH built their Quebec DC, but I'd be surprised if their expansion plans don't include leasing space (if they don't already do it).
If you count edge locations (where stuff like anycasted nameservers, cloudfront and network PoPs are hosted) as well then the DC count is significantly higher than that
Don't forget the availability zones! Each region has 2-4 availability zones, and each of these has 1-6 datacenters. The scale of AWS is at least an order of magnitude more important than OVH's.
It's not uncommon to use multiple providers, that way you don't need to rely on a single provider having multiple DC's in every location you want a presence.
Amazon has a peered connection at KNIX, although from most locations ingress traffic to the Seoul region is routed through AWS Tokyo first, so latency is probably going to be slightly worse than Tokyo unless accessed from within Korea.
Pricing seems very similar if not identical to the Tokyo region, which is not surprising since bandwidth is rather expensive in SK. Overall I feel this is built to satisfy SK's data privacy laws rather than commercial reasons.
I imagine you were getting downvoted for using multiple question marks, and people not understanding you were just asking a question.
Typically one question mark indicates a question. Multiple question marks usually indicates that you're upset about something or asking a rhetorical question.
For example, most people distinguish a difference between sentences like:
"What are you doing?"
and
"What are you doing???"
The first indicates a question, the second indicates you are upset or that the person is doing something wrong.
The way you wrote your sentence makes it look like you could be making fun of the person you are responding to.
Censored internet access though (Korean Porn-Blocker)
[ec2-user@ec2 ~]$ curl youporn.com
<html><script>
var arg = "http://warning.or.kr";
var str = new Array();
str = arg.split("&", 1);
var a = new Array();
a = str[0].split("=");
var b = Math.floor(a[1] / 100);
var c = new Array();
if(b == 10){location.replace("http://www.naver.com");}
else if(b == 20){location.replace("http://www.daum.net");}
else if(b == 30){location.replace("http://www.paran.com");}
else{ c = a[0].split("?");
location.replace(c[0]);}
</script></html>
I'd argue that m3 and c3 instances shouldn't be classified as "previous generation", since the m4 and c4 instances don't have any local ephemeral storage. Given the recent EBS incident in GovCloud, I think it's still pretty reasonable to be skeptical of EBS.
