Response requested, Google de-peered from NWAX
Amber Pham
NWAX member survey regarding Google de-peering
TL;DR
Google has recently “de-peered” from NWAX, meaning that their ASN is no longer reachable on the NWAX switching fabric. The NWAX Board would like to understand how concerned you are, and whether you think something should be done in response.
Please use this survey form to let us know: https://forms.gle/aeuAVLgbxZ6tYAan6
The full explanation:
To read this online: Click Here
Google has recently “de-peered” from NWAX, meaning that their ASN is no longer reachable on the NWAX switching fabric. They have expressed that they have a strategy of deactivating “remote” IX locations, and despite having some of their largest facilities in the Dalles, NWAX is considered a remote IX.
Explanations from Google via NANOG talk: https://www.youtube.com/watch?v=Yg-qV6Fktjw&ab_channel=NANOG
We recognize that NWAX members are interested in the content that Google provides. Google’s choice to de-peer creates technical challenges for traffic engineering, potentially leading to increased costs on transit links. This situation may also necessitate alternative interconnection strategies for our members that could be costly, difficult, or even impossible to implement.
The NWAX board has considered several different options to provide a partial or complete solution to this problem. Some may have direct costs, some may have indirect costs. We have put together a list of the various options below, and are seeking insight from our member community to determine which path we should pursue.
There are three methods that we see as potential paths forward for NWAX:
NWAX could take on the costs of re-connecting to Google, and create a “transit ASN” whose only function would be to reach Google prefixes. This ASN would then appear on NWAX, but would be selectively peered only with NWAX members who paid for bilateral interconnection to this ASN to offset the costs of the downstream connection.
Pros:
Simple configuration options
Requires no significant architecture changes for members
Cons:
Subsidizes Google at member expense for traffic that was local/zero cost
Rewards de-peering trend
Incurs additional costs for members who want to participate
Incurs significant additional complexity for NWAX for architecture/billing
Google is still peering at the SIX in Seattle. NWAX could extend to Seattle and inter-connect to the SIX and provide a VLAN-based method for NWAX members to optionally interconnect with ALL SIX peers including Google. NWAX would effectively be operating a SIX extension as a secondary benefit for NWAX members.
Pros:
Embodies “Northwest Access Exchange” between major markets in Oregon and Washington. Not just a fix for connecting to Google.
Zero-cost interconnection model is maintained.
Possible longer-term options for a SIX relationship.
360+ Additional peers accessible to NWAX members - https://www.seattleix.net/participants
Expands potential membership base for additional IX members directly connecting to NWAX switch in Seattle
Reasonable cost impact to NWAX, transport is the only significant recurring cost.
Cons:
Google is not guaranteed to remain peering at SIX forever.
Long-haul fiber still has a cost to NWAX
Latency to peers in Seattle is ~5 milliseconds. The close proximity might discourage content providers from building infrastructure in Oregon.
Google previously connected to NWAX from Seattle, so latency to Google would be approximately the same as it was before.
Might discourage content providers from building infrastructure in Oregon.
VLANs add minor complexity to member configuration (NWAX already provides trunked VLAN’s to members for our Jumbo frame fabric, the additional SIX VLAN would be a seamless config change, members can choose to use it or ignore it)
Reaches capacity fairly quickly due to long-haul cost/fiber speed (400G should last quite a while given that it currently exceeds the bandwidth utilization of the entire IX). Dark fiber would allow us to expand our bandwidth almost indefinitely with only one time hardware costs.
Do nothing.
Pros:
Costs nothing
Does not reward de-peering behaviors
Cons:
Shifts Google traffic to transit links for members; costs more
Reduces value proposition for NWAX
Please use this survey form: https://forms.gle/aeuAVLgbxZ6tYAan6
