[Internal Instance] - Globally available world-id-indexer

[paolodamico]

To have extremely low latency on the world-id-indexer, considering this deployment structure.

• Deploy full indexer instances in 3x k8s clusters in three different regions (Virginia, Frankfurt, Singapore).
• Use AWS Global Accelerator or DNS routing (Route53) to route traffic to the lowest latency region (healthy) for the user.

[asurak]

Who/what is going to request those services? Do you want to do "best region" selection on the infrastructure side based on some network assumptions or rather do a light client-side logic for selection? Client-side selection would also better support overrides because of particular preferences.

[asurak]

Another potential benefit of client-side selection is to allow greater decentralization. AWS Global Accelerator works only for AWS upstreams. If the client selects, we can eventually mix infrastructure and CDN/WAF providers. Today we have AWS+Cloudflare and we could do GCP+Fastly for some of the regions.

[paolodamico]

Either option works well. For OPRF nodes, the client will pick a region based on their device timezone. Based on the decentralization argument, it seems reasonable to me that we proceed in this direction. CC @danielllek