Over the past few years, the popularity of cloud applications has rapidly increased, requiring a thorough understanding of cloud workloads to continue improving the state of the art. To facilitate studying cloud workloads, we built a benchmark suite of three network-intensive applications - a Web 2.0 social networking site, a video-streaming server, and a key-value store. In this work, we present the challenges of benchmarking cloud workloads and demonstrate the use of our benchmark suite through two case studies that compare the behavior of these cloud applications on a 32-core MIPS processor and a traditional 4-core x86 CPU.