Switchboard Crossbar Hosting

Multi-region infrastructure deployment for Switchboard Crossbar instances

Table of Contents

• Overview
• Sponsored Endpoints
  • Endpoint Usage
  • Adding Your Own Endpoints
  • 📢 Stay Updated
• Basic Design and Architecture
• Future Work
• Steps
• Caveats
• Contributing

Overview

Basic repo that allows you to run your own crossbar instance very simply. The main motivations:

• The crossbar application is very performant and scaling based on resources is not the bottleneck.
• Surge connects to many exchanges. If a single outgoing IP address is making too many requests, the exchanges will rate limit the IP address. This repo is built with that in mind.
• The closer you are to Tokyo the lower the latency.

Sponsored Endpoints

Company	Region	Endpoint	Stability	Notes
P0	asia-northeast2 (Seoul)	https://34.97.218.183.sslip.io	🟢 Production	Primary endpoint, regional coverage
P0	asia-northeast1 (Tokyo)	https://35.200.6.110.sslip.io	🟡 Staging	Staging endpoint, high availability but used to test new versions

Note: All endpoints use preemptible nodes for cost optimization. Minor downtime (seconds) may occur during node preemption, but this saves 60-90% on infrastructure costs.

Endpoint Usage

• Primary: Use asia-northeast3 for best performance
• Load Balancing: Rotate between endpoints to avoid rate limits
• Testing: Use the stage environment for development

Adding Your Own Endpoints

Want to sponsor additional endpoints? Deploy using this repo and submit a PR with your company logo and endpoint details!

📢 Stay Updated

Join our Telegram channel for endpoint updates, maintenance notifications, and community discussions: @switchboard_endpoints

Basic Design and Architecture

• Each region gets its own isolated network, cluster, node pool, and most importantly a regional static IP address.
• If you want two nodes, you simply define two regions. This means two separate outgoing IP addresses. This means you can make twice as many requests without being rate limited.
• Use preemptible nodes to save money. This means you might have some minor downtime (seconds) if a node is preempted. But it can save you 60-90% on node costs.
• This uses stable tag. It's best to hardcode a version.

Future Work

• Global Load Balancer: This might be nice instead of having clients rotate the URL's they use for thier requests.
• API Key: If users want to secure their instance they can use an API key. Although for early implementation its not necessary.
• Multi-Cloud: Add integrations for AWS, Azure, and any other cloud provider.

Steps

1. Install pulumi and setup all your credentials.. Start from cd pulumi.
2. cp env.example stage.env and cp env.example prod.env
3. Init your stack: pulumi stack init {{org}}/stage and pulumi stack init {{org}}/prod respectively.
4. Select your stack: pulumi stack select {{org}}/stage or pulumi stack select {{org}}/prod. It's recommended to have two environemnts. This way you can test new crossbar tags without affecting the production environment.
5. Run pulumi up to deploy the infrastructure.
6. If you run into issues, you can run pulumi up again in most cases. Other times, you might need to pulumi destroy and then pulumi up again.
7. Once the infrastructure is deployed, you can see all your new IPs. They look something like this: https://xx.xxx.xx.xx.sslip.io

Caveats

1. SSL certs can take some time to get ready. Use this to test: curl -k "https://{{your_ip}}.sslip.io/simulate/solana/mainnet/EAsoLo2uSvBDx3a5grqzfqBMg5RqpJVHRtXmjsFEc4LL?includeReceipts=true"
2. If you want to destroy. First you need to set the deletionProtection to false in the Pulumi.dev.yaml file.
   1. Then you do a pulumi up to update the state.
   2. Then you can do a pulumi destroy to destroy the infrastructure.
3. If your certs don't work, delete them: kubectl delete certificate crossbar-us-{region}-tls-secret --context=gke_{project_name}_{region}_crossbar-{region}-cluster

Contributing

Contributions are welcome. Simply create an issue first, we can discuss the changes you want to make. If you want to make a change, you can fork the repo and create a pull request.