Volume 18, No. 5, 2021
Dependency Aware Caching (Dac) For Software Defined Networks
V. Rajkumar , Dr. V. Maniraj
Abstract
By utilising software-defined networking (SDN), control applications are granted the ability to establish fine-grained forwarding policies within the switches that sit beneath them. As a direct result of this, the number of wildcard rule patterns that hardware switches are able to implement is restricted due to the price and energy requirements of Ternary Content Addressable Memory (also known as terabytes) (TCAM). To add insult to injury, the hardware switches in question are unable to manage a quick series of rule table updates, thereby compounding the problem. In this investigation, we demonstrate how to combine the most advantageous aspects of hardware and software processing in order to provide programmes the impression of high-speed forwarding, big rule tables, and quick updates. This strategy "caches" the rules that are used the most frequently in a very tiny TCAM, and it relies on software to regulate the relatively little amount of "cache miss" traffic that is produced. Cache-replacement strategies that are now available cannot be implemented in a blind method, however, because of the interdependencies that exist between different rules that share similar patterns. It is advised to "splice" big dependency chains in order to cache smaller sets of rules while still keeping the policy's semantics. This may be done by splitting up the chains into many smaller chains. Experiments that were carried out with our DAC prototype suggested that rule splicing may be an effective method for utilising limited TCAM space while simultaneously reacting quickly to changes in both policy and traffic.
Pages: 2403-2412
Keywords: SDN, Caching, dependency aware caching, TCAM.