Same here, most of the AWS related network loads very slow.

In-fact it is worse today.

l43a2:

 

a ton of sites that use aws in sydney are slow, bigpipes own site, lightbox, had to give a buddy access to an au VPN to play PUBG tonight

 

yup thanks for that vpn ;) 

Theres also a discussion thread about it on Spark's FB page: https://www.facebook.com/spark4nz/posts/1663307550355744

The packet loss from Global Gateway across Telstra Global is absolutely killer. 

A lot of services that use Amazon Cloudfront are unusable. Even using my EC2 instances just via SSH is stuttery 

I've found using Mosh to Amazon works well in replacement to SSH but it is still pretty bad.

Hopefully Spark fixes this soon. You don't understand how many services are hosted at AWS until even one of their datacentres are affected.

yitz: It's the path traffic is taking on the way back that is affected, they had direct peering with Amazon as recently as 2 months ago, not sure what's changed.

possibly fibre cut or the likes, this is upstream and on the reverse so although in progress on being resolved, a little outside of flicking a switch control on our end.

FYI, according to a BigPipe response on their FB page : "It appears the the troubles are with how the upstream traffic is being routed from GGI to AWS and the engineers on both ends are still looking into things."

It won't be an easy fix - routing never is. Thanks for all the techs that are working on it.

yitz: Better now?

certainly does seem to be mitigated.  

yitz: Better now?

My 300Mb download from Cloudfront that took about 8 hours to fetch 200Mb and now has just finished the last 100Mb in 15 seconds or so would point to yes :)

Local AWS guys on Twitter confirmed it's been fixed and is all back to normal.

As usual their support is _awesome_. Can't say enough good things about the local AWS peeps.

(Not that BigPipe were bad when I asked them.... )

Yep seeing routing is improved and my speeds (whilst slower than normal) are back on par. Good job to everyone involved in a fix :)

looking good here