Hey FD, let's talk AWS hosting

[stimutacs]

OK, so, with the recent "no offline mode" issue, this is somewhat more important.

Here's what I regularly receive from the hop before your hosting server, depending on the time of day and which governments/hackers are DDOS whoever - typically when it's bad, it'll go

14 997 ms 844 ms 1118 ms ec2-79-125-0-137.eu-west-1.compute.amazonaws.com
[79.125.0.137]

to

14 3359 ms 3477 ms 3279 ms ec2-79-125-0-73.eu-west-1.compute.amazonaws.com
[79.125.0.73]

kinds of figures.

It also suffers from occasional loss, even when it is behaving at regular 350-390ms latency (and this is with an annex M low latency profile, seriously about as good as this will ever get from Brisbane au without fibre) - typically in the range of 2% or so over 100, but can go higher than this depending on load. People elsewhere in Australia are also going to have a 4000km addition over a not especially great link, to what is already basically a trip longer than the entire circumference of the earth, bringing them to 400-500+ms.

Amazon Elastic Compute Cloud claim they have several compute service options nearby - eg -

http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-regions-availability-zones.html

Code Name

ap-northeast-1 Asia Pacific (Tokyo)

ap-southeast-1 Asia Pacific (Singapore)

ap-southeast-2 Asia Pacific (Sydney)

eu-central-1 EU (Frankfurt)

eu-west-1 EU (Ireland)

sa-east-1 South America (Sao Paulo)

us-east-1 US East (N. Virginia)

us-west-1 US West (N. California)

us-west-2 US West (Oregon)

Are these servers appropriate for running ED's back end? Do you plan to run the back end from other AWS locations? If so, is there a timeline for bringing this online?

 

[Michael Brookes]

The simple answer is yes they are suitable. The quick response requirements are all peer to peer and we try to group players geographically to minimise response time. For the main servers it doesn't cause a problem for the game if the response take a few more milliseconds.

Michael

 

[Seboss]

Besides, connectivity between AWS VPCs located in different regions is a pain in the butt

 

[stimutacs]

The simple answer is yes they are suitable. The quick response requirements are all peer to peer and we try to group players geographically to minimise response time. For the main servers it doesn't cause a problem for the game if the response take a few more milliseconds.

Michael

So ... I guess there is no timeline? Because while 350ms is borderline usable - picking the right time of day and your luck - it really isn't when it stretches to multiple seconds, or, worse case, both multiple seconds lag and loss.

Besides, connectivity between AWS VPCs located in different regions is a pain in the butt

Right. I mean - can I just mention that I am really yet to be sold on the benefits of the living universe as a concept? I mean - taking out the idea of even trying to sync between servers for a moment - I seem to be trading off critical connection issues for the ability to buy gold for 999 space bucks rather than 1000 space bucks depending on whether people I can neither see nor care about are doing a thing? Whee?