<![CDATA[Message Queuing Telemetry Transport (MQTT) is a publish / subscribe client communication protocol with the message exchange process. MQTT depends on TCP / IP as its base layer (Dinesh Thangavel, 2014) so the MQTT protocol work in realtime. In real time condition the broker must ensure all connections at the same time. However, a large number of clients connected to the broker will result in increased load on the broker. An increase in load can affect existing resources such as network bandwidth, memory, CPU usage, and throughput to overcome this problem can use a load balancing mechanism. Round robin algorithms can be used for load balancing but would be less efficient when applied to systems that use a broker with a different devices specification. Therefore, it is necessary to implement other load balancing algorithms at the MQTT broker on devices that have different specifications. The weighted least connection algorithm can be used to solve these problems because it can work by determining the weight of each broker. The parameters taken from this test include the number of connections, response time, and CPU usage. Tests are carried out with several scenarios, namely scenarios during the publish process and during the publish-subscribe process, with a predetermined amount. The result of this research is that the application of load balancing can reduce CPU usage when handling the publish-subscribe process with 1000 clients.]]>
Copyrights © 2021
