Hi all, a nanog thread started on November 23, 2018 discussed the challenges of getting Amazon peering sessions turned up. Has anyone had luck since/does anyone have a contact they could refer me to — off-list or otherwise? The process of getting PNI in place with other CSPs was straightforward, but I haven’t heard back from AWS after a month and several follow-ups. Our customers would really benefit from us getting this sorted.
Thanks,
Kelly
There are many folks that here that are in AWS. Assuming you have followed what is in https://aws.amazon.com/peering/ (and https://aws.amazon.com/peering/policy/) then send me details privately about what/when/who and I’ll reach out internally to the relevant folks.
Have gotten into the habit of making annual peering requests to Amazon asking turn up a session on a shared IXP peering. Once was able to get a peering session turned up, no traffic was ever shifted onto it before we moved out of that carrier hotel a year or so later. The amazon peering email box does have humans surfing it.
Over the years a number of network operators have mentioned getting little response from Amazon about peering requests.
For a company like Amazon they have little reason to do peering with small scale operators. They already peer with the tier 1’s and assume I will do what I need to balance my bits. The fancy algorithms they use to balance traffic around does allow them to operate a decent network with fewer staff and less links to the small ISPs. Just a network operator here, trying to get my bytes across the wire.
Enjoy your weekend!
Have gotten into the habit of making annual peering requests to Amazon asking turn up a session on a shared IXP peering. Once was able to get a peering session turned up, no traffic was ever shifted onto it before we moved out of that carrier hotel a year or so later. The amazon peering email box does have humans surfing it.
Over the years a number of network operators have mentioned getting little response from Amazon about peering requests.
For a company like Amazon they have little reason to do peering with small scale operators. They already peer with the tier 1’s and assume I will do what I need to balance my bits. The fancy algorithms they use to balance traffic around does allow them to operate a decent network with fewer staff and less links to the small ISPs. Just a network operator here, trying to get my bytes across the wire.
Enjoy your weekend!
There are also three different ASNs and different policy decision trees; they all report selective criteria such as minimum of 10GE, multiple locations, specific locations, etc. Not sure it’s as simple as ‘getting the right person’ more than it is about meeting the right conditions. Easier for network operators. Enterprises may be different (and better off with their upstreams or PacketFabric).
YMMV,
-M<
Hi all,
We (Amazon) have made improvements over the last few years to significantly improve our NOC and Peering response times, as well as the time it takes to set up a peering session, whether over IX or direct PNI/SFI.
If you don’t get a response from the respective Peering mailing address listed on our PeeringDB https://peeringdb.com/net/1418 , feel free to email me privately and I’ll get in touch with the right folks internally.
Regards,
Andras
For those persons who have not received an answer from the Amazon peering email addresses, or a BGP session with traffic actually flowing across it…
Obviously Amazon does not share their own traffic volume criteria for selecting a peer vs. sending traffic to them over a giant IP transit provider.
I wonder what the actual threshold is as measured in traffic volume from netflow data to/from the Amazon AS before they start taking a potential peer seriously. Obviously if you’re somebody big like a regional ILEC or a cable operator that has half of a major city as your incumbent territory, it’s not even a question, but for smaller ISPs it’s an interesting question to discover where exactly that threshold is.
I think everyone is missing the point.
There are lots of companies that “peer” with AWS and Azure or GCP. That’s just easier for latency and for ensuring the traffic doesn’t greatly travel over the public internet if they have a interconnect at a POP we also happen to be in.
But as a consequence of charging for egress and not for ingress, they have built their own global network on the same levels as Cogent and Level3. In fact, their network engineers would joke that if my mickey mouse company of a internal “wan” ever broke, we could conceivably just route over their network and some of our resiliency plans take that into account.
If you have kept up with the research and testing they put out in an annual report, the network observability company Thousand Eyes does an amazing job and excellent writeup regarding how each cloud provider does.
It dives deep into how out of all the cloud service providers, AWS was one of the worse in latency and congestion due to other two major cloud providers having went that “extra mile” (pun intended).
As a result, they are putting their network edge “closer” to their customer base vs Amazon who makes the packets travel to one of their datacenters or a “Mega-POP” and at that point they are on that networks faster 10GigE (or more).
But no, there’s no reason for AWS to do a traditional peering as if they were an ISP, they aren’t. They’re more like a black hole for or a volcano of bytes that enter the internet at large. However, they will peer with anyone for a small interconnect charge. It’s all MPLS. 
Here is the URL for the white paper on their benchmark should you not want to add another row to to their spam generator.
Findings start at page 14 with the previous describing their the testing and methodology.