Internet caches serve a dual purpose. First, caches minimize access latencies for objects that are shared by multiple clients or accessed repeatedly by a single client. Second, caches improve performance for all users by reducing load on the network backbone and servers. While private browser caches can filter repeated accesses, only shared caches can eliminate the need for a site to deliver a given object to every client that requests it, helping to slow traffic growth and absorb bursts in demand for popular sites.
CRISP allows construction of very large distributed proxy caches to serve the needs of Internet-connected organizations (ISPs, commercial or otherwise). This is important because demand-side caches are a critical part of the solution for scalability problems in the Internet. ISPs have the incentive and resources to deploy caches to improve service to their members or customers. In this way, the cost of expanded capacity is borne by those who benefit from it the most.
We have argued that CRISP's simple implementation of cooperative caching, based on a central mapping service, is effective and efficient for distributed Web caches serving tens of thousands of users. In particular, the properties of Internet object access (read-only access to relatively large objects with small URLs, tolerance of latencies below human perception, and the ability to fall back to the Internet if the map fails) allow us to safely enjoy the simplicity and benefits of the centralized structure.