Downloads: 4 | Views: 68 | Weekly Hits: ⮙1 | Monthly Hits: ⮙1
Technical Report | Information Technology | United States of America | Volume 13 Issue 2, February 2024
Strategizing Storage Deployment in Compute Centric Public Clouds: Balancing Capacity, Congestion, and Cost for Evolving Workloads
Anurag Reddy [2] | Anil Naik [2] | Sandeep Reddy [5]
Abstract: The paper explores the intricacies of deploying storage services in an existing compute-only environment within public clouds, considering factors such as capacity expansion, traffic congestion risks, and cost implications. The focus is on aligning data center architecture to support evolving workloads including real-time high bandwidth, low latency applications like AR/VR gaming, immersive streaming, health care (Telemedicine), military applications & monitoring systems while meeting performance, latency, reliability, and cost standards. The introduction emphasizes the importance of adapting data center architecture to seamlessly accommodate diverse workloads in public clouds. The paper outlines options for deploying storage services, aiming for reasonable and predictable lead times for capacity expansions. Key variables, including long-term forecasting, are introduced, considering high-density compute servers and forecasts for both compute and storage servers for the next year. Four design scenarios are presented, each with its benefits and risks. Design 1 proposes high storage density mixed-racks, highlighting quick deployment with no additional costs but raises concerns about potential traffic congestion. Design 2 suggests low storage density mixed-racks to handle recovery traffic but incurs incremental hardware costs and longer deployment times. Designs 3 and 4 focus on storage servers-only racks and spines, offering advantages like quick deployment and isolation from compute resources during rebuilds but come with higher failure domain risks. In conclusion, the paper underscores the need for aligning network architecture with long-term forecasts for scalability and flexibility. It emphasizes optimizing designs to balance performance, reliability, and cost considerations. The presented scenarios offer a comprehensive exploration of challenges and opportunities in deploying storage services within a compute-centric environment, providing insights for data center strategists and architects.
Keywords: Data center architecture, compute-only environment, and storage services deployment with a focus on capacity expansion, traffic congestion risks, low latency, cost implications in public clouds
Edition: Volume 13 Issue 2, February 2024,
Pages: 1355 - 1358