|  | 
 
| I've always been curious about how these large-scale stations achieve real-time consistency across nodes. Are there any projects that can do this? I want to make some idle servers available. If we use multiple-node backend, then when CDN pulls back-end nodes, how should we choose the fastest one? It seems that either random importance weights or round-robin are used in the CDN pull algorithm. This could potentially slow down.
 This is what I am looking for.
 This is just an example and not what I am aiming for.
 | 
 |