A good start for you is the wiki page on Multi-agent systems:

https://en.wikipedia.org/wiki/Multi-agent_system

A major advantage of using multiple agents is that different tasks, capabilities and behaviours can be assigned to different sets of agents or particles.

Some consequent drawbacks are that the program will be more complicated, probably slower, and more prone to human error. In some cases it might also be very difficult to understand why the system as a whole is working well, or why it is failing to achieve the goals that have been set for it.

Not sure if I understand your question correctly, but the distributed nature of the algorithm is actually a good reason to use MAS: if you state a distributed algorithm as a MAS, It is easier to implement the decoupling, to understand its function, or to tune the parameters.

Multi-agent systems run on the distributed environment (Sometimes in a single computer) capable to communicate with each other through the message parsing. Through the communication, agents can solve complex problems easily. If you use a correct framework it is not probably slower. However, algorithm processes are strata forward. Note that MAS has the different approach than the algorithmic process. In addition, MAS can provide results that cannot previously predict. Try it.

@Budditha Hettige: thank you for your answer, i wanted to ask if running a Multi-agent system in a SINGLE computer will increase the CPU time?

In multi-agent system there are possibilities for agents to coordinate their actions, as well as to use stochastic methods in decision-making.