Load Balancer

• Periodic Healthechks are performed for the status of the Server
  • HealthCheck interval can be specified in the config.json
  • status_counter indicates the maximum number of heath checks to be performed once a server is down.
  • pingInterval indicates that max amount of time the load-balancer waits for a pong from the server
• Load Balancing Strategies
  • cpumetrics, random, roundrobin, weightedroundrobin can be specified as the load balancing strategy in config.json
  • cpumetrics are collected in load-balancer by invoking the API /systemResources exposed by the server for every pingInterval time
  • Information about new servers can be added at runtime by invoking the API /server

    POST  /server
    Host: load-balancer
    Body: 
    {
    	"serverName": "server3",
    	"address": "localhost:8084",
    	"weight": 2
    }
    

  • If a server goes down, details about it are automically updated in load-balancer and request are not routed to it.
• Status Completion of Continuous Queries
  • API /status is exposed in the load-balancer to mark the completion of the continuous queries by the servers. Continuous queries of type /resources are only remembered by the load-balancer (to configure them via config.json can be implemented).

  POST /status
  Host: load-balancer
  Body: 
  {
     "backend" : "localhost:8081",
     "starttime": "Tue, 06/06/17, 08:19AM",
     "endtime": "Wed, 06/07/17, 10:50AM"
  }
  

• Dynamic Routing of the continuous queries /resources is handled when a processing node goes down.

Team Members : Sri Krishna, Thejdeep, Raghu