The Experts below are selected from a list of 135 Experts worldwide ranked by ideXlab platform
Kun Wang - One of the best experts on this subject based on the ideXlab platform.
-
HPCA - Power shifting in Thrifty Interconnection Network
2011 IEEE 17th International Symposium on High Performance Computer Architecture, 2011Co-Authors: Jian Li, Wei Huang, Charles Lefurgy, Lixin Zhang, Wolfgang E. Denzel, Richard R. Treumann, Kun WangAbstract:This paper presents two complementary techniques to manage the power consumption of large-scale systems with a packet-switched interconnection Network. First, we propose Thrifty Interconnection Network (TIN), where the Network links are activated and de-activated dynamically with little or no overhead by using inherent system events to timely trigger link activation or de-activation. Second, we propose Network Power Shifting (NPS) that dynamically shifts the power budget between the compute nodes and their corresponding Network components. TIN activates and trains the links in the interconnection Network, just-in-time before the Network communication is about to happen, and thriftily puts them into a low-power mode when communication is finished, hence reducing unnecessary Network power consumption. Furthermore, the compute nodes can absorb the extra power budget shifted from its Attached Network components and increase their processor frequency for higher performance with NPS. Our simulation results on a set of real-world workload traces show that TIN can achieve on average 60% Network power reduction, with the support of only one low-power mode. When NPS is enabled, the two together can achieve 12% application performance improvement and 13% overall system energy reduction. Further performance improvement is possible if the compute nodes can speed up more and fully utilize the extra power budget reinvested from the thrifty Network with more aggressive cooling support.
-
Power shifting in Thrifty Interconnection Network
2011 IEEE 17th International Symposium on High Performance Computer Architecture, 2011Co-Authors: Jian Li, Wei Huang, Charles Lefurgy, Lixin Zhang, Wolfgang E. Denzel, Richard R. Treumann, Kun WangAbstract:This paper presents two complementary techniques to manage the power consumption of large-scale systems with a packet-switched interconnection Network. First, we propose Thrifty Interconnection Network (TIN), where the Network links are activated and de-activated dynamically with little or no overhead by using inherent system events to timely trigger link activation or de-activation. Second, we propose Network Power Shifting (NPS) that dynamically shifts the power budget between the compute nodes and their corresponding Network components. TIN activates and trains the links in the interconnection Network, just-in-time before the Network communication is about to happen, and thriftily puts them into a low-power mode when communication is finished, hence reducing unnecessary Network power consumption. Furthermore, the compute nodes can absorb the extra power budget shifted from its Attached Network components and increase their processor frequency for higher performance with NPS. Our simulation results on a set of real-world workload traces show that TIN can achieve on average 60% Network power reduction, with the support of only one low-power mode. When NPS is enabled, the two together can achieve 12% application performance improvement and 13% overall system energy reduction. Further performance improvement is possible if the compute nodes can speed up more and fully utilize the extra power budget reinvested from the thrifty Network with more aggressive cooling support.
A. Mcdonald - One of the best experts on this subject based on the ideXlab platform.
-
Hot Interconnects - Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
-
Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
A. Gavrilovska - One of the best experts on this subject based on the ideXlab platform.
-
Splits stream handlers: deploying application-level services to Attached Network processors
2020Co-Authors: A. Gavrilovska, K. SchwanAbstract:Modern distributed applications utilize a rich variety of distributed services. Due to the computation-centric notions of modern machines, application-level implementations of these services are problematic for applications requiring high data transfer rates, for reasons that include the inability of modern architectures to efficiently execute computations with communication. Conversely, Network-level implementations of services are limited due to the Network's inability to interpret application-level data or execute application-level operations on such data. The emergence of programmable Network processors capable of high-rate data transfers, with flexible interfaces for external reconfiguration, has created new possibilities for movement of processing into the Network infrastructure. This thesis explores the extent to which programmable Network processors can be used in conjunction with standard host nodes, to form enhanced computational host-ANP (Attached Network Processor) platforms that can deliver increased efficiency for variety of applications and services. The main contributions of this research are the creation of SPLITS, a Software architecture for Programmable LIghtweighT Stream handling, dynamic configuration of data paths through the host-ANP nodes, and the dynamic creation, deployment and reconfiguration of application-level processing applied along these paths. With SPLITS, application-specific services can be dynamically mapped to the host, ANP, or both, to best exploit their joint capabilities. activities are stream handlers—parameterizable, lightweight, computation units that operate on data headers as well as application-level content. Experimental results demonstrate performance gains of executing various application-level services on ANPs, and demonstrate the importance of the SPLITS host-ANP nodes to support dynamically reconfigurable services, and to deal with the resource limitations on the ANPs.
-
Hot Interconnects - Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
-
Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
Jian Li - One of the best experts on this subject based on the ideXlab platform.
-
HPCA - Power shifting in Thrifty Interconnection Network
2011 IEEE 17th International Symposium on High Performance Computer Architecture, 2011Co-Authors: Jian Li, Wei Huang, Charles Lefurgy, Lixin Zhang, Wolfgang E. Denzel, Richard R. Treumann, Kun WangAbstract:This paper presents two complementary techniques to manage the power consumption of large-scale systems with a packet-switched interconnection Network. First, we propose Thrifty Interconnection Network (TIN), where the Network links are activated and de-activated dynamically with little or no overhead by using inherent system events to timely trigger link activation or de-activation. Second, we propose Network Power Shifting (NPS) that dynamically shifts the power budget between the compute nodes and their corresponding Network components. TIN activates and trains the links in the interconnection Network, just-in-time before the Network communication is about to happen, and thriftily puts them into a low-power mode when communication is finished, hence reducing unnecessary Network power consumption. Furthermore, the compute nodes can absorb the extra power budget shifted from its Attached Network components and increase their processor frequency for higher performance with NPS. Our simulation results on a set of real-world workload traces show that TIN can achieve on average 60% Network power reduction, with the support of only one low-power mode. When NPS is enabled, the two together can achieve 12% application performance improvement and 13% overall system energy reduction. Further performance improvement is possible if the compute nodes can speed up more and fully utilize the extra power budget reinvested from the thrifty Network with more aggressive cooling support.
-
Power shifting in Thrifty Interconnection Network
2011 IEEE 17th International Symposium on High Performance Computer Architecture, 2011Co-Authors: Jian Li, Wei Huang, Charles Lefurgy, Lixin Zhang, Wolfgang E. Denzel, Richard R. Treumann, Kun WangAbstract:This paper presents two complementary techniques to manage the power consumption of large-scale systems with a packet-switched interconnection Network. First, we propose Thrifty Interconnection Network (TIN), where the Network links are activated and de-activated dynamically with little or no overhead by using inherent system events to timely trigger link activation or de-activation. Second, we propose Network Power Shifting (NPS) that dynamically shifts the power budget between the compute nodes and their corresponding Network components. TIN activates and trains the links in the interconnection Network, just-in-time before the Network communication is about to happen, and thriftily puts them into a low-power mode when communication is finished, hence reducing unnecessary Network power consumption. Furthermore, the compute nodes can absorb the extra power budget shifted from its Attached Network components and increase their processor frequency for higher performance with NPS. Our simulation results on a set of real-world workload traces show that TIN can achieve on average 60% Network power reduction, with the support of only one low-power mode. When NPS is enabled, the two together can achieve 12% application performance improvement and 13% overall system energy reduction. Further performance improvement is possible if the compute nodes can speed up more and fully utilize the extra power budget reinvested from the thrifty Network with more aggressive cooling support.
K. Schwan - One of the best experts on this subject based on the ideXlab platform.
-
Splits stream handlers: deploying application-level services to Attached Network processors
2020Co-Authors: A. Gavrilovska, K. SchwanAbstract:Modern distributed applications utilize a rich variety of distributed services. Due to the computation-centric notions of modern machines, application-level implementations of these services are problematic for applications requiring high data transfer rates, for reasons that include the inability of modern architectures to efficiently execute computations with communication. Conversely, Network-level implementations of services are limited due to the Network's inability to interpret application-level data or execute application-level operations on such data. The emergence of programmable Network processors capable of high-rate data transfers, with flexible interfaces for external reconfiguration, has created new possibilities for movement of processing into the Network infrastructure. This thesis explores the extent to which programmable Network processors can be used in conjunction with standard host nodes, to form enhanced computational host-ANP (Attached Network Processor) platforms that can deliver increased efficiency for variety of applications and services. The main contributions of this research are the creation of SPLITS, a Software architecture for Programmable LIghtweighT Stream handling, dynamic configuration of data paths through the host-ANP nodes, and the dynamic creation, deployment and reconfiguration of application-level processing applied along these paths. With SPLITS, application-specific services can be dynamically mapped to the host, ANP, or both, to best exploit their joint capabilities. activities are stream handlers—parameterizable, lightweight, computation units that operate on data headers as well as application-level content. Experimental results demonstrate performance gains of executing various application-level services on ANPs, and demonstrate the importance of the SPLITS host-ANP nodes to support dynamically reconfigurable services, and to deal with the resource limitations on the ANPs.
-
Hot Interconnects - Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
-
Stream handlers: application-specific message services on Attached Network processors
Proceedings 10th Symposium on High Performance Interconnects, 2002Co-Authors: A. Gavrilovska, K. Mackenzie, K. Schwan, A. McdonaldAbstract:This paper presents a software architecture that enables the application-specific processing of messages on Network processors Attached to cluster machines. Such processing is performed by stream handlers executed on such Attached Network processors (ANPs) and able to manipulate both message headers and their data content. Handler execution can be associated with the ANPs' receive side, its transmit side, or both. Using Intel's IXP1200 boards as sample ANPs, the paper evaluates performance advantages and tradeoffs in stream handler execution. Results indicate that while receive-side stream customization is useful for simple stream handlers, it becomes a bottleneck and results in degraded performance with increased handler complexity or with increased amounts of data manipulated by handlers. In comparison, transmit-side handler execution exhibits more constant and acceptable performance levels and can therefore, support richer ANP functionality.
