Hello Jaipal,
OSPF and IS-IS are both link state protocols.
IS-IS is lighter in cpu resources and can be more easy to update (to add new capabilities on the protocol for example IPv6 support was added in an easier way by using TLV type, length, value data structures).
Thanks to TLV approach IS-IS update and flooding process is more simple: OSPF deals with multiple data structures of different types with different propagation characteristics).
So IS-IS is supposed to be more scalable.
OSPF provides more support for NBMA networks (it has more network types then IS-IS) and a is able to calculate automatically costs of interfaces.
IS-IS sees IP subnets as leafs in its CLNS topology so it has been the first to introduce and use partial recalculation where a subset of SFP is recalculated.
One drawback of IS-IS is that its packets are not IPv4 and so they cannot be prioritized with current modular QoS.
US based providers usually use IS-IS.
In Europe OSPF is used widely.
For example my current customer that is a big enterprise uses IS-IS. before I worked on big service providers networks that were using and are using OSPF.
A very big service provider network may use BGP confederations and multiple IGPs domains.
With MPLS and iBGP widely deployed the IGP databases are reduced to minimum size so scalability is not a very big concern.
What I can say with day by day experience with IS-IS is that sometimes I see IS-IS adjacencies to drop on links that are not sature and I think this comes from the fact that modular QoS leaves them in the default class.
Hope to help
Giuseppe
