Dubbo

1. RPC (Socket + serialization)+ cloud

   Node role: Consumer of the Provider service Invokes the Consumer of the remote service Registry Registry of the service registration and discovery Monitor Collects statistics about service invocation times and invocation time and sends statistics to the monitoring center Container every minute Service run container (started at the service provider)Copy the code

   Load balancing:1.Random load balancing based on weights (% chance)2.Polling load balancing mechanism based on weight3.Least active load balancing mechanism (who is the fastest to find who)4.Consistent load balancing mechanismCopy the code

High availability: Service providers and consumers can still communicate through local caches even after all of dubbo's exposed service registries are downCopy the code

Service downgraded: according to the actual situation and the traffic, for some services have a strategy to stop or in processing, simple ways to release the server resources to ensure the normal operation of the core business This is to prevent the distributed service avalanche effect (when there is a high request timeout under many concurrent requests have been know downtime service resource depletion, After the outage, other distributed services will invoke the outage service resources are exhausted, and the outage occurs. The entire distributed service is down. This is an avalanche.1.Configure service degradation on the console: masking (consumer methods on the service are returned directlynullValue) and fault tolerance (return on call failurenullValue)Copy the code