Why we need virtual chassis in a network ? What is the use of it ? Why not have one switch instead of having 2 or more switches (Virtual chassis) in a network .
There are two main reasons for using stacking techniques like virtual chassis: redundancy and ease of management.
If you have just a single switch and it fails in some way, then your network is completely down. If you have a stack of switches and one node in the stack fails, then you have some form of connectivity left. When you design your network right, it can survive failure of any single device in the network.
As for the second reason: if you want to provide access for 400 endpoints, you would need a very large switch to provide for all those access ports. More likely, you would use 10 smaller switches because that's a lot cheaper usually. When you stack those 10 switches, you can manage them as a single device, making your job a lot easier.
Some reasons I haven't seen mentioned yet:
ease of cabling: instead of having hundreds of cables which need to go to one specific switch in one rack, you can use a chassis consisting of a number of switches, which gives you (if you have enough rack space) more room to do decent cable management
pay as you grow: you may not need hundreds of ports directly, but you may need them in a year time. Buying more small switches which become part of a virtual chassis when you need them enables you to spread costs in time
Further to RobinG's answer - the single control plane in virtual chassis (and other stacking technologies) allows you to do thinks like attach an aggregated ethernet link between the virtual chassis and a downstream switch, and distribute the member links across both chassis nodes.
If you had two stand-alone switches uplinked to a downstream switch, this would not work, and you'd have to resort to active/standy links using Spanning-Tree or similar (or MC-LAG if your switch supports it)
Virtual chassis' are an improvement over chassis based switches primarily because of MLAG support.
Redundancy improvement is true only insofar as passive metal is a point of failure. This is a matter of probability and is quote low.
Scalability is not significantly improved. In fact, 48-port linecards are often cheaper than 48 port switches. So, as you get very large chassis, the cost per port actually decreases.
Management is improved only insofar as a campus or org-wide virtual chassis is deployed.