These are the following benefits:

1- REDUCED LATENCY

Compared to a device-to-cloud architecture, placing processing closer to the devices can reduce the latency since the physical distance is shorter and potential response time in a data center can be removed. Compared to a device only architecture, latency can be reduced since computation intensive tasks that take a long time on resource-constrained sensor devices can be moved to more capable fog computing nodes. The motivation can also be to keep the latency predictable.

2- PRIVACY

Compared to the device-to-cloud architecture, fog computing can reduce the propagation of data, for instance by analyzing sensitive data on a local gateway instead of a data center outside of the control of the user. This can improve the privacy of user data.

3- ENERGY EFFICIENCY

There are several ways how fog computing can improve energy efficiency within sensor devices. First, gateways can serve as communication proxies, so that devices can increase the length of their sleep cycles. During the sleep mode, the gateway takes care of any requests or updates, which are then processed when the sensor device wakes up. Second, energy-intensive computations and other services can be offloaded from the battery-driven nodes.

4- BANDWIDTH

In comparison to a device-to-cloud architecture, fog computing can reduce the volume of data to be sent into data centers. This can happen in several ways: Raw data can be filtered, analyzed, pre-processed or compressed so that only a reduced amount of data needs to be forwarded. Local nodes can also answer requests from devices based on locally cached data, so that communication with data centers is not necessary at all.

I agree with Rahul Yadav and this article is explaining more about it:

Article Fog Computing and its Role in the Internet of Things

disadvantages will be increasing cost and in some times it also adds some complexity to your ecosystem.