At the application level, both the workload mix and usage patterns have evolved, with increased focus on business applications and emphasis on service-centric computing and SLA-driven performance tuning.
At the business level, there are increased cost pressures from commoditization as well as increased competitive pressures as vendors of servers, storage, services and software all seek to position themselves in the new enterprise landscape.
These dramatic changes in the enterprise IT landscape motivate equivalent changes in the emphasis of computer systems architecture research. In particular, we believe that system architecture design has to be revisited in the context of the interactions with the overall ecosystem.
For example, individual systems need to be optimized for their role in the broader data center leveraging cross connections across the hardware/software boundaries, and with a focus on total cost of ownership. In addition, the new solutions have to be consistent with current trends toward leveraging industry-standard and commodity components.
Our goal is to design a commodity-based, high-volume compute fabric for the next-generation data center. This fabric will be customizable and focus on delivering customer value beyond performance on aspects such as manageability, power efficiency, availability, security, etc., as differentiators at a minimal cost.
Our research contributes to the state-of-the-art in computer systems at three primary levels:
The structure and emphasis of enterprise IT systems has changed considerably over the past decade. Large, consolidated data centers have emerged. Virtualization, multi-core processors and modularization are growing in popularity. Hardware is increasingly commoditized.
This work addresses end-user value along multiple dimensions - performance, power, manageability and availability. The blades++ research is part of HP's broader initiative on the next-generation data center.
- Multicores and virtualization
- Platform design for next-generation blade and interconnect architectures
- Data center architectures
- "Enterprise IT trends and implications on system architecture research," Parthasarathy Ranganathan and Norman Jouppi, Proceedings of the International Conference on High-Performance Computer Architecture (HPCA), 2005
- "Heterogeneous chip multiprocessors," Rakesh Kumar, Dean Tullsen, Norman Jouppi, Parthasarathy Ranganathan, IEEE Computer, November 2005 (cover feature)
- "Configurable Isolation: Building High Availability Systems with Commodity Multi-Core Processors," Nidhi Aggarwal, Parthasarathy Ranganathan, Norman Jouppi, James Smith, Proceedings of the International Symposium on Computer Architecture (ISCA), June 2007
- "Ensemble-level Power Management for Dense Blade Servers," Parthasarathy Ranganathan, Phil Leech, David Irwin, and Jeff Chase, Proceedings of the International Symposium on Computer Architecture (ISCA), June 2006
- "General-Purpose Blade Infrastructure for Configurable System Architectures," Kevin Leigh, Parthasarathy Ranganathan, Jaspal Subhlok, Journal on Parallel and Distributed Databases (JPDD), Springer Publishers 0926-8782 (Print) 1573-7578, March 2007
- "Making Scheduling Cool: Temperature-aware Resource Scheduling," Justin Moore, Jeff Chase, Parthasarathy Ranganathan, Ratnesh Sharma. Proceedings of the 2005 Annual Usenix Conference (USENIX), April 2005.
- Smart Data Center, Chandrakant Patel et al., Proceedings of HP TechCon, April 2003
- "No Power Struggles: A Unified Power Management Architecture for the Data Center," Phil Leech, Parthasarathy Ranganathan, Vanish Talwar, Tom Vaden, Xiaoyun Zhu, and Zhikui Wang, Proceedings of HP Techcon, April 2007
- "Simulating Complex Enterprise Workloads using Utilization Traces," Parthasarathy Ranganathan and Philip Leech, Tenth Workshop on Computer Architecture Evaluation using Commercial Workloads (CAECW), held with HPCA-13, February 2007
- "Data Center Workload Monitoring, Analysis, and Emulation," Justin Moore, Jeff Chase, Keith Farkas, and Parthasarathy Ranganathan. In the Eighth Workshop on Computer Architecture Evaluation using Commercial Workloads (CAECW), February, 2005. (Invited paper)