We will use a random write workload that uniformly writes 4KB pages over the Full Backend that is available. As long as the size of the cache is relatively smaller than the Full Backend size, the workload has little temporal locality, and will produce nearly 100% write misses. The think time, namely, the pause between completing one write request and issuing another one, of this workload is set to zero. In other words, this workload is capable, in principle, of driving a storage system at an infinite throughput. The throughput is limited only by the capacity of the system to serve the writes. This workload is extremely helpful in profiling behavior of the algorithms across a wide range of NVS sizes.