We implemented a framework, called WorkGen as shown in Figure 5, which allows us to benchmark various algorithms across a large range of cache sizes, varying page sizes, and consumption rates. WorkGen is divided into three layers. The first layer consists of the workload modules, which can simulate any number of parallel streams consisting of sequential, short sequential, SPC1-Read, or SPC2-VOD workloads. The second layer consists of the prefetching modules, which allows us to select any of the prefetching algorithms that we implemented. Each algorithm module has access to the LRU data structure as well as the read(), and prefetch() methods used to perform demand reads and prefetches respectively. The third layer consists of the disk backend which is the target for all the I/Os.
