Course Description

This course covers the state-of-the-art hardware and software architecture of datacenters that power almost all the cloud services we use today. We will discuss the specific requirements for datacenter hardware and software architecture and how these requirements influence datacenter design. We will explore new trends in processor architecture (both general-purpose and domain-specific), memory architecture, storage architecture, network architecture, system software architecture, and datacenter software architecture.

Topics to Cover (Tentative)

• Datacenter Introduction and History
  • Google case study
• Datacenter Software
  • Software layers, killer applications, development styles
• Datacenter Performance
  • Metrics, tail latency, profiling
• Datacenter Energy and Power
  • Metrics, energy proportionality, power management
• Datacenter Storage and Memory
  • Memory and storage disaggregation, tiered memory
• Datacenter Networks
  • Topologies, control plane, data plane
• Datacenter Compute
  • Post-Moore CPUs, accelerators
• Novel Paradigms
  • 2.5D and 3D packaging, near-data processing