NaaS: Network-as-a-Service in the Cloud Computing

This research paper proposes Network-as-a-Service (NaaS) as an advanced cloud computing model that gives tenants direct, secure, and programmable access to data-center network infrastructure, overcoming the limitations of traditional clouds where users control compute and storage but not networking. By enabling network visibility, custom packet forwarding, and in-network processing, NaaS allows applications to implement services such as efficient multicast, load balancing, data aggregation, caching, redundancy elimination, and application-specific routing directly within switches and routers rather than only at end hosts. The paper introduces a NaaS architecture featuring programmable “NaaS boxes” running in-network processing elements (INPEs), along with APIs and a phased processing model to ensure scalability, concurrency, and efficient resource use. Through simulations on a large fat-tree data-center topology, the study demonstrates that even modest processing capabilities can significantly reduce network traffic and flow completion time while improving overall throughput for both NaaS-aware and traditional applications. The work concludes that NaaS is feasible on current hardware and has the potential to enhance performance, flexibility, and efficiency in cloud environments, although challenges such as scalability, isolation, programmability, and pricing models remain for widespread adoption.