In moments of high load is there concerns with the load balancer being the bottleneck of the system. How exactly are load balancers handled to make sure that even with a high amount of requests the system doesn't suffer in latency. Is the general idea to make more load balancers in order to reduce the bottleneck but increase overhead or to improve the load balancer or some combination of both?
bojianh
One thing about adding more load balancer is that there may be a need to perform some sort of synchronization between the load balancers which can also be costly. But as seen in Assignment4, the master node's load balancer does very little work and spends very little time on each task relative to the task itself.
In moments of high load is there concerns with the load balancer being the bottleneck of the system. How exactly are load balancers handled to make sure that even with a high amount of requests the system doesn't suffer in latency. Is the general idea to make more load balancers in order to reduce the bottleneck but increase overhead or to improve the load balancer or some combination of both?
One thing about adding more load balancer is that there may be a need to perform some sort of synchronization between the load balancers which can also be costly. But as seen in Assignment4, the master node's load balancer does very little work and spends very little time on each task relative to the task itself.